Pd-dev November 2012

pd-dev@lists.iem.at

9 participants
83 discussions

[ pure-data-Bugs-3588893 ] [autoabstraction] not working in PdX
by SourceForge.net 21 Nov '12

21 Nov '12

Bugs item #3588893, was opened at 2012-11-21 02:24 Message generated for change (Tracker Item Submitted) made by zmoelnig You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=478070&aid=3588893&group_… Please note that this message will contain a full copy of the comment thread, including the initial issue submission, for this request, not just the latest update. Category: pd-extended Group: v0.43 Status: Open Resolution: None Priority: 5 Private: No Submitted By: IOhannes m zmölnig (zmoelnig) Assigned to: Hans-Christoph Steiner (eighthave) Summary: [autoabstraction] not working in PdX Initial Comment: iemgut's "autoabstraction" feature is not working in recent Pd-extended builds. expected behaviour: instantiating [autoabstraction] will try to load a template and complain if that fails. on success the template can be used for creating abstractions on the fly observed behaviour: instantiating [autoabstraction] works, but doesn't print anything, regardless of verbosity and whether the template-file exists or not. after loading, the automatic abstraction creation does not work at all. platform: OSX-10.? (macintel), Pd-exteneded downloaded via puredata.info/pd-extended/beta on 2012-11-15 (i don't have access to the machines no more, and i think the download links were changed yesterday) ---------------------------------------------------------------------- You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=478070&aid=3588893&group_…

1 0

[ pure-data-Patches-3588876 ] NGPWhDsIiWqhCoM
by SourceForge.net 21 Nov '12

21 Nov '12

Patches item #3588876, was opened at 2012-11-21 00:07 Message generated for change (Tracker Item Submitted) made by nobody You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=478072&aid=3588876&group_… Please note that this message will contain a full copy of the comment thread, including the initial issue submission, for this request, not just the latest update. Category: pd-extended Group: feature Status: Open Resolution: None Priority: 5 Private: No Submitted By: Nobody/Anonymous (nobody) Assigned to: Nobody/Anonymous (nobody) Summary: NGPWhDsIiWqhCoM Initial Comment: kamagra 100 ---------------------------------------------------------------------- You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=478072&aid=3588876&group_…

1 0

canvasdelete search method
by Jonathan Wilkes 21 Nov '12

21 Nov '12

I see canvasdelete adds a "delete" method for canvas. I'm thinking of adding a "deleteany" method that takes A_GIMME and deletes any object that matches the args. Examples: [deleteany foo(--[sendcanvas] deletes [foo] and [foo 12] [deleteany foo 12(--[sendcanvas] deletes [foo 12] but not [foo] Or maybe this interface would fit better with dynamic patching: [deleteany obj foo(--[sendcanvas] deletes [foo] and [foo 12] but not [foo( [deleteany text foo 12(--[sendcanvas] deletes comment "foo 12" [deleteany msg(--[sendcanvas] would delete all message boxes Basically combining the core functionality of the "find" method with the "delete" method. It would be a mildly useful improvement, but could be very useful if there were something like this: [connectany "obj foo" 0 "obj bar" 0(--[sendcanvas] [foo] [bar] becomes [foo]--[bar] Unfortunately Pd doesn't have sublists, and symbols with spaces are still pretty clunky. Maybe connectany should be a separate external rather than a canvas method. Can anyone think of a decent interface for something like this? It would vastly improve the flexibility of dynamic patching. -Jonathan

2 3

[ pure-data-Patches-3588794 ] hHkJlURkPlVCkXimve
by SourceForge.net 20 Nov '12

20 Nov '12

Patches item #3588794, was opened at 2012-11-20 12:54 Message generated for change (Tracker Item Submitted) made by nobody You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=478072&aid=3588794&group_… Please note that this message will contain a full copy of the comment thread, including the initial issue submission, for this request, not just the latest update. Category: externals Group: documentation Status: Open Resolution: None Priority: 5 Private: No Submitted By: Nobody/Anonymous (nobody) Assigned to: Nobody/Anonymous (nobody) Summary: hHkJlURkPlVCkXimve Initial Comment: apcalis oral jelly buy ---------------------------------------------------------------------- You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=478072&aid=3588794&group_…

1 0

[ pure-data-Patches-3587384 ] Replacing deprecated functions
by SourceForge.net 20 Nov '12

20 Nov '12

Patches item #3587384, was opened at 2012-11-14 21:06 Message generated for change (Comment added) made by eighthave You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=478072&aid=3587384&group_… Please note that this message will contain a full copy of the comment thread, including the initial issue submission, for this request, not just the latest update. >Category: puredata Group: bugfix Status: Open Resolution: None Priority: 5 Private: No Submitted By: Nobody/Anonymous (nobody) >Assigned to: Miller Puckette (millerpuckette) Summary: Replacing deprecated functions Initial Comment: vexp_fun.c currently uses the old BSD-style functions drem and finite. These functions have been removed from XCode 4.4 and later, so that the expr~ family of externals fails to compile now. This patch replaces drem with remainder and finite with isfinite, which should work everywhere. ---------------------------------------------------------------------- >Comment By: Hans-Christoph Steiner (eighthave) Date: 2012-11-20 12:34 Message: Makes sense to me. remainder() is the C99 replacement for drem(), and isfinite() is the C99 replacement for finite(). C99 marks drem() and finite() as deprecated. ---------------------------------------------------------------------- You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=478072&aid=3587384&group_…

1 0

[ pure-data-Patches-3581737 ] OpenBSD support
by SourceForge.net 20 Nov '12

20 Nov '12

Patches item #3581737, was opened at 2012-10-29 16:36 Message generated for change (Comment added) made by eighthave You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=478072&aid=3581737&group_… Please note that this message will contain a full copy of the comment thread, including the initial issue submission, for this request, not just the latest update. Category: puredata Group: None Status: Open Resolution: None Priority: 5 Private: No Submitted By: Federico Schwindt (fgsch) Assigned to: Nobody/Anonymous (nobody) Summary: OpenBSD support Initial Comment: FIrst round of patches to add support for OpenBSD. ---------------------------------------------------------------------- >Comment By: Hans-Christoph Steiner (eighthave) Date: 2012-11-20 12:00 Message: that looks good to me, Miller, but I don't know the BSD stuff. ---------------------------------------------------------------------- Comment By: Miller Puckette (millerpuckette) Date: 2012-11-20 09:08 Message: I think HAVE_ALLOCA_H is traditionally used to indicate the existence of a file, "alloca.h" which apparently doesn't exist in MSVC. I'm no longer worried about how we can get teh makefiles set up but rather what should go in the C code... at teh moment I believe it #ifdef HAVE_ALLOCA_H # include <alloca.h> /* linux, mac, mingw, cygwin */ #elif defined _MSC_VER # include <malloc.h> /* MSVC */ #else # include <stddef.h> /* BSDs for example */ #endif ... and that I can put that in, clean out the "old" autoconf stuff and voila. Do we all think that that work OK? ---------------------------------------------------------------------- Comment By: IOhannes m zmölnig (zmoelnig) Date: 2012-11-20 08:51 Message: or don't use config.h at all, add /D=HAVE_ALLOCA_H to makefile.nt (for those compiling on w32 using msvc) and use AC_CONFIG_HEADERS(alloca.h) for the systems that use autoconf. i doubt whether anybody involved in writing configure.ac has done the testing on all platforms known to compile Pd before submitting changes. ---------------------------------------------------------------------- Comment By: Hans-Christoph Steiner (eighthave) Date: 2012-11-19 19:46 Message: if you really think its worth your time, then go ahead and prepare a patch that you've tested on OpenBSD, GNU/Linux, Windows, Mac OS X, iOS, Android, Rockbox, and Microsoft Visual Studio. IMHO, things are working well as they are. ---------------------------------------------------------------------- Comment By: Federico Schwindt (fgsch) Date: 2012-11-19 15:01 Message: Not necessarily. You can supply pre-generated config.h for some platforms (that are obviously not going to change) and generate one for the rest. So basically if you're using autoconf you could use AC_CONFIG_HEADERS, if not just supply a config.h with all the defines for that platform. ---------------------------------------------------------------------- Comment By: Hans-Christoph Steiner (eighthave) Date: 2012-11-19 11:02 Message: AC_CONFIG_HEADERS works fine if your code is only built by the single autotools build system. In Pd's case, there are multiple build systems because of all of the platforms it runs on. There is configure.ac for UNIX and MinGW, src/makefile.nt for Microsoft Visual Studio, libpd for Android, iOS and more. Pd is also included in Rockbox. So using some auto-tools things makes it harder on the other platforms. ---------------------------------------------------------------------- Comment By: Miller Puckette (millerpuckette) Date: 2012-11-19 08:03 Message: I'm preparing to throw out the old "configure" stuff (which, by the way, I didn't originate and which I've tried and failed to understand; for some reason I never got AC_CONFIG_HEADERS() to work). But I'd still like to know what a good way to declare alloca(0 in a C file is. The autoconf web site suggests that I cut and past a 16-line #ifdef mess into every c file using alloca() .. perhaps under the "new" (autogen) system something shorter would work, but I'd want to have some confidence I'm doing something reasonable. ---------------------------------------------------------------------- Comment By: Federico Schwindt (fgsch) Date: 2012-11-17 19:37 Message: unfortunately you're wrong. alloca is not posix so i wouldn't expect any agreement here. as i said openbsd doesn't have alloca.h. i would not be surprised if other bsd are the same. malloc.h is also incorrect. here alloca's prototype is in stdlib.h. i will prepare a diff to add the maze of defines in all the configures and what not but I really fail to see why you can simply use AC_CONFIG_HEADERS() since you're already depending on configure. ---------------------------------------------------------------------- Comment By: Miller Puckette (millerpuckette) Date: 2012-11-17 15:03 Message: I went ahead and applied the patch by hand. I could be wrong, but isn't there now an alloca() function and an alloca.h on all unix-likes? and an alloca() defined in Windows? If so we can blow off all the configuration stuff and just conditionall include alloca or malloc.h depending on whether we're Windows. ---------------------------------------------------------------------- Comment By: IOhannes m zmölnig (zmoelnig) Date: 2012-11-13 15:45 Message: no not really. afaiu the plan is to add a check for "alloca.h" to configure, which willl define HAVE_ALLOCA_H if the file is found. if HAVE_ALLOCA_H is defined alloca.h will get included. now the current behaviour is to include alloca.h in any case. but if the include is protected, then the build-system has to actively define HAVE_ALLOCA_H in order keep the old behaviour. simply ignoring the other build-systems might break them. ---------------------------------------------------------------------- Comment By: Hans-Christoph Steiner (eighthave) Date: 2012-11-13 11:45 Message: I think you'll be safe only touching the ./configure.ac and Makefile.ams. Miller seems to want to hold only to the old, crufty src/configure.in build, but he doesn't use OpenBSD. ---------------------------------------------------------------------- Comment By: IOhannes m zmölnig (zmoelnig) Date: 2012-11-06 05:13 Message: #1 yes, the build-system is messed up (or rather, it is in a state of transition) #2 .../src/configure adds HAVE_foo to CPPFLAGS (and .../configure adds it to DEFS), which get expanded in the resp. makefiles (so technically you are right, they don't get added to CFLAGS) #3 i said "makefile.nt" and not "makefile.in", which is the 3rd(!) build-system that is there and which is used to build on w32 using MSVC. ---------------------------------------------------------------------- Comment By: Nobody/Anonymous (nobody) Date: 2012-11-06 04:45 Message: I might be missing something but unless you add it manually I don't see where configures adds the HAVE_XXX to CFLAGS unless you do it manually. And src/makefile.in is processed by src/configure(.in) so only ../configure.ac and ../src/configure.in need to be modified. <rant> You have a pretty messed up build system. </rant> ---------------------------------------------------------------------- Comment By: IOhannes m zmölnig (zmoelnig) Date: 2012-11-06 03:57 Message: if the patch comes from git, then please create a branch, commit to that and then create a patch-set using "git format-patch master". as for the HAVE_ALLOCA_H: if you don't instruct configure to write defines into config.h, it will add them to the CFLAGS, so that the compilation will still work. thus, no need to change the build-system. bear in mind though, that currently Pd uses to separate build-systems: .../configure.ac and .../src/configure.in (and .../src/makefile.nt); you might need to modify all 3 in order to get accepted (yes, it's a PITA) ---------------------------------------------------------------------- Comment By: Federico Schwindt (fgsch) Date: 2012-11-06 02:48 Message: @zmoelnig: The patch comes from git. My main concern is how to proceed with the other changes that require more changes than just extending an existing ifdef. For example the exclusion of alloca.h, etc. In my git clone I've modified configure to use config headers and so I can do ifdef HAVE_ALLOCA_H, etc. but I'm not sure if such patches will be accepted. The reason I've break them down is so they can be committed while I understand what your preference is. ---------------------------------------------------------------------- Comment By: IOhannes m zmölnig (zmoelnig) Date: 2012-11-06 02:39 Message: this patch is simple enough (almost trivial) as it only adds "#ifdef __OpenBSD__" to the clauses that are already handled by "__FreeBSD__". i don't see a reason to not include it (but of course this all depends on miller). small note on patches: - the patch doesn't apply here (`patch` thinks the diff is garbage) - Pd is developped in git; it would be great if you could generate your patches using git as well; see [1] - as for atomicity of patches: personally i think creating a (few) smaller patches is not a real problem; make sure that those patches are logical units rather than reflecting your iterative trial-and-error history to fix a problem. then attach all patch files to this ticket (rather than creating a new patchtracker-ticket for each file) [1] http://puredata.info/docs/developer/GitWorkflows/ ---------------------------------------------------------------------- Comment By: Federico Schwindt (fgsch) Date: 2012-11-06 02:03 Message: Comments please? I'd really to move forward with this. Thanks. ---------------------------------------------------------------------- Comment By: Federico Schwindt (fgsch) Date: 2012-10-29 19:01 Message: That'd be ideal but since I'm not sure about the other changes (using config headers, etc.) I'm submitting what I know for certain that won't change depending on the approach. For example what should I do if alloca.h doesn't exist? In my git repo I have #ifdef HAVE_ALLOCA_H where the define is driven from autotools header check but I'm still unclear whether you want to use the headers or not (and the reason you might not want to do it since you're already using autotools). Happy to submit everything in one go if you can suggest what the best course of action is. ---------------------------------------------------------------------- Comment By: Hans-Christoph Steiner (eighthave) Date: 2012-10-29 18:57 Message: The best thing to do would be to make a single patch that includes everything needed for OpenBSD support. ---------------------------------------------------------------------- You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=478072&aid=3581737&group_…

1 0

[ pure-data-Patches-3581737 ] OpenBSD support
by SourceForge.net 20 Nov '12

20 Nov '12

Patches item #3581737, was opened at 2012-10-29 16:36 Message generated for change (Comment added) made by millerpuckette You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=478072&aid=3581737&group_… Please note that this message will contain a full copy of the comment thread, including the initial issue submission, for this request, not just the latest update. Category: puredata Group: None Status: Open Resolution: None Priority: 5 Private: No Submitted By: Federico Schwindt (fgsch) Assigned to: Nobody/Anonymous (nobody) Summary: OpenBSD support Initial Comment: FIrst round of patches to add support for OpenBSD. ---------------------------------------------------------------------- >Comment By: Miller Puckette (millerpuckette) Date: 2012-11-20 09:08 Message: I think HAVE_ALLOCA_H is traditionally used to indicate the existence of a file, "alloca.h" which apparently doesn't exist in MSVC. I'm no longer worried about how we can get teh makefiles set up but rather what should go in the C code... at teh moment I believe it #ifdef HAVE_ALLOCA_H # include <alloca.h> /* linux, mac, mingw, cygwin */ #elif defined _MSC_VER # include <malloc.h> /* MSVC */ #else # include <stddef.h> /* BSDs for example */ #endif ... and that I can put that in, clean out the "old" autoconf stuff and voila. Do we all think that that work OK? ---------------------------------------------------------------------- Comment By: IOhannes m zmölnig (zmoelnig) Date: 2012-11-20 08:51 Message: or don't use config.h at all, add /D=HAVE_ALLOCA_H to makefile.nt (for those compiling on w32 using msvc) and use AC_CONFIG_HEADERS(alloca.h) for the systems that use autoconf. i doubt whether anybody involved in writing configure.ac has done the testing on all platforms known to compile Pd before submitting changes. ---------------------------------------------------------------------- Comment By: Hans-Christoph Steiner (eighthave) Date: 2012-11-19 19:46 Message: if you really think its worth your time, then go ahead and prepare a patch that you've tested on OpenBSD, GNU/Linux, Windows, Mac OS X, iOS, Android, Rockbox, and Microsoft Visual Studio. IMHO, things are working well as they are. ---------------------------------------------------------------------- Comment By: Federico Schwindt (fgsch) Date: 2012-11-19 15:01 Message: Not necessarily. You can supply pre-generated config.h for some platforms (that are obviously not going to change) and generate one for the rest. So basically if you're using autoconf you could use AC_CONFIG_HEADERS, if not just supply a config.h with all the defines for that platform. ---------------------------------------------------------------------- Comment By: Hans-Christoph Steiner (eighthave) Date: 2012-11-19 11:02 Message: AC_CONFIG_HEADERS works fine if your code is only built by the single autotools build system. In Pd's case, there are multiple build systems because of all of the platforms it runs on. There is configure.ac for UNIX and MinGW, src/makefile.nt for Microsoft Visual Studio, libpd for Android, iOS and more. Pd is also included in Rockbox. So using some auto-tools things makes it harder on the other platforms. ---------------------------------------------------------------------- Comment By: Miller Puckette (millerpuckette) Date: 2012-11-19 08:03 Message: I'm preparing to throw out the old "configure" stuff (which, by the way, I didn't originate and which I've tried and failed to understand; for some reason I never got AC_CONFIG_HEADERS() to work). But I'd still like to know what a good way to declare alloca(0 in a C file is. The autoconf web site suggests that I cut and past a 16-line #ifdef mess into every c file using alloca() .. perhaps under the "new" (autogen) system something shorter would work, but I'd want to have some confidence I'm doing something reasonable. ---------------------------------------------------------------------- Comment By: Federico Schwindt (fgsch) Date: 2012-11-17 19:37 Message: unfortunately you're wrong. alloca is not posix so i wouldn't expect any agreement here. as i said openbsd doesn't have alloca.h. i would not be surprised if other bsd are the same. malloc.h is also incorrect. here alloca's prototype is in stdlib.h. i will prepare a diff to add the maze of defines in all the configures and what not but I really fail to see why you can simply use AC_CONFIG_HEADERS() since you're already depending on configure. ---------------------------------------------------------------------- Comment By: Miller Puckette (millerpuckette) Date: 2012-11-17 15:03 Message: I went ahead and applied the patch by hand. I could be wrong, but isn't there now an alloca() function and an alloca.h on all unix-likes? and an alloca() defined in Windows? If so we can blow off all the configuration stuff and just conditionall include alloca or malloc.h depending on whether we're Windows. ---------------------------------------------------------------------- Comment By: IOhannes m zmölnig (zmoelnig) Date: 2012-11-13 15:45 Message: no not really. afaiu the plan is to add a check for "alloca.h" to configure, which willl define HAVE_ALLOCA_H if the file is found. if HAVE_ALLOCA_H is defined alloca.h will get included. now the current behaviour is to include alloca.h in any case. but if the include is protected, then the build-system has to actively define HAVE_ALLOCA_H in order keep the old behaviour. simply ignoring the other build-systems might break them. ---------------------------------------------------------------------- Comment By: Hans-Christoph Steiner (eighthave) Date: 2012-11-13 11:45 Message: I think you'll be safe only touching the ./configure.ac and Makefile.ams. Miller seems to want to hold only to the old, crufty src/configure.in build, but he doesn't use OpenBSD. ---------------------------------------------------------------------- Comment By: IOhannes m zmölnig (zmoelnig) Date: 2012-11-06 05:13 Message: #1 yes, the build-system is messed up (or rather, it is in a state of transition) #2 .../src/configure adds HAVE_foo to CPPFLAGS (and .../configure adds it to DEFS), which get expanded in the resp. makefiles (so technically you are right, they don't get added to CFLAGS) #3 i said "makefile.nt" and not "makefile.in", which is the 3rd(!) build-system that is there and which is used to build on w32 using MSVC. ---------------------------------------------------------------------- Comment By: Nobody/Anonymous (nobody) Date: 2012-11-06 04:45 Message: I might be missing something but unless you add it manually I don't see where configures adds the HAVE_XXX to CFLAGS unless you do it manually. And src/makefile.in is processed by src/configure(.in) so only ../configure.ac and ../src/configure.in need to be modified. <rant> You have a pretty messed up build system. </rant> ---------------------------------------------------------------------- Comment By: IOhannes m zmölnig (zmoelnig) Date: 2012-11-06 03:57 Message: if the patch comes from git, then please create a branch, commit to that and then create a patch-set using "git format-patch master". as for the HAVE_ALLOCA_H: if you don't instruct configure to write defines into config.h, it will add them to the CFLAGS, so that the compilation will still work. thus, no need to change the build-system. bear in mind though, that currently Pd uses to separate build-systems: .../configure.ac and .../src/configure.in (and .../src/makefile.nt); you might need to modify all 3 in order to get accepted (yes, it's a PITA) ---------------------------------------------------------------------- Comment By: Federico Schwindt (fgsch) Date: 2012-11-06 02:48 Message: @zmoelnig: The patch comes from git. My main concern is how to proceed with the other changes that require more changes than just extending an existing ifdef. For example the exclusion of alloca.h, etc. In my git clone I've modified configure to use config headers and so I can do ifdef HAVE_ALLOCA_H, etc. but I'm not sure if such patches will be accepted. The reason I've break them down is so they can be committed while I understand what your preference is. ---------------------------------------------------------------------- Comment By: IOhannes m zmölnig (zmoelnig) Date: 2012-11-06 02:39 Message: this patch is simple enough (almost trivial) as it only adds "#ifdef __OpenBSD__" to the clauses that are already handled by "__FreeBSD__". i don't see a reason to not include it (but of course this all depends on miller). small note on patches: - the patch doesn't apply here (`patch` thinks the diff is garbage) - Pd is developped in git; it would be great if you could generate your patches using git as well; see [1] - as for atomicity of patches: personally i think creating a (few) smaller patches is not a real problem; make sure that those patches are logical units rather than reflecting your iterative trial-and-error history to fix a problem. then attach all patch files to this ticket (rather than creating a new patchtracker-ticket for each file) [1] http://puredata.info/docs/developer/GitWorkflows/ ---------------------------------------------------------------------- Comment By: Federico Schwindt (fgsch) Date: 2012-11-06 02:03 Message: Comments please? I'd really to move forward with this. Thanks. ---------------------------------------------------------------------- Comment By: Federico Schwindt (fgsch) Date: 2012-10-29 19:01 Message: That'd be ideal but since I'm not sure about the other changes (using config headers, etc.) I'm submitting what I know for certain that won't change depending on the approach. For example what should I do if alloca.h doesn't exist? In my git repo I have #ifdef HAVE_ALLOCA_H where the define is driven from autotools header check but I'm still unclear whether you want to use the headers or not (and the reason you might not want to do it since you're already using autotools). Happy to submit everything in one go if you can suggest what the best course of action is. ---------------------------------------------------------------------- Comment By: Hans-Christoph Steiner (eighthave) Date: 2012-10-29 18:57 Message: The best thing to do would be to make a single patch that includes everything needed for OpenBSD support. ---------------------------------------------------------------------- You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=478072&aid=3581737&group_…

1 0

[ pure-data-Patches-3581737 ] OpenBSD support
by SourceForge.net 20 Nov '12

20 Nov '12

Patches item #3581737, was opened at 2012-10-29 16:36 Message generated for change (Comment added) made by zmoelnig You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=478072&aid=3581737&group_… Please note that this message will contain a full copy of the comment thread, including the initial issue submission, for this request, not just the latest update. Category: puredata Group: None Status: Open Resolution: None Priority: 5 Private: No Submitted By: Federico Schwindt (fgsch) Assigned to: Nobody/Anonymous (nobody) Summary: OpenBSD support Initial Comment: FIrst round of patches to add support for OpenBSD. ---------------------------------------------------------------------- >Comment By: IOhannes m zmölnig (zmoelnig) Date: 2012-11-20 08:51 Message: or don't use config.h at all, add /D=HAVE_ALLOCA_H to makefile.nt (for those compiling on w32 using msvc) and use AC_CONFIG_HEADERS(alloca.h) for the systems that use autoconf. i doubt whether anybody involved in writing configure.ac has done the testing on all platforms known to compile Pd before submitting changes. ---------------------------------------------------------------------- Comment By: Hans-Christoph Steiner (eighthave) Date: 2012-11-19 19:46 Message: if you really think its worth your time, then go ahead and prepare a patch that you've tested on OpenBSD, GNU/Linux, Windows, Mac OS X, iOS, Android, Rockbox, and Microsoft Visual Studio. IMHO, things are working well as they are. ---------------------------------------------------------------------- Comment By: Federico Schwindt (fgsch) Date: 2012-11-19 15:01 Message: Not necessarily. You can supply pre-generated config.h for some platforms (that are obviously not going to change) and generate one for the rest. So basically if you're using autoconf you could use AC_CONFIG_HEADERS, if not just supply a config.h with all the defines for that platform. ---------------------------------------------------------------------- Comment By: Hans-Christoph Steiner (eighthave) Date: 2012-11-19 11:02 Message: AC_CONFIG_HEADERS works fine if your code is only built by the single autotools build system. In Pd's case, there are multiple build systems because of all of the platforms it runs on. There is configure.ac for UNIX and MinGW, src/makefile.nt for Microsoft Visual Studio, libpd for Android, iOS and more. Pd is also included in Rockbox. So using some auto-tools things makes it harder on the other platforms. ---------------------------------------------------------------------- Comment By: Miller Puckette (millerpuckette) Date: 2012-11-19 08:03 Message: I'm preparing to throw out the old "configure" stuff (which, by the way, I didn't originate and which I've tried and failed to understand; for some reason I never got AC_CONFIG_HEADERS() to work). But I'd still like to know what a good way to declare alloca(0 in a C file is. The autoconf web site suggests that I cut and past a 16-line #ifdef mess into every c file using alloca() .. perhaps under the "new" (autogen) system something shorter would work, but I'd want to have some confidence I'm doing something reasonable. ---------------------------------------------------------------------- Comment By: Federico Schwindt (fgsch) Date: 2012-11-17 19:37 Message: unfortunately you're wrong. alloca is not posix so i wouldn't expect any agreement here. as i said openbsd doesn't have alloca.h. i would not be surprised if other bsd are the same. malloc.h is also incorrect. here alloca's prototype is in stdlib.h. i will prepare a diff to add the maze of defines in all the configures and what not but I really fail to see why you can simply use AC_CONFIG_HEADERS() since you're already depending on configure. ---------------------------------------------------------------------- Comment By: Miller Puckette (millerpuckette) Date: 2012-11-17 15:03 Message: I went ahead and applied the patch by hand. I could be wrong, but isn't there now an alloca() function and an alloca.h on all unix-likes? and an alloca() defined in Windows? If so we can blow off all the configuration stuff and just conditionall include alloca or malloc.h depending on whether we're Windows. ---------------------------------------------------------------------- Comment By: IOhannes m zmölnig (zmoelnig) Date: 2012-11-13 15:45 Message: no not really. afaiu the plan is to add a check for "alloca.h" to configure, which willl define HAVE_ALLOCA_H if the file is found. if HAVE_ALLOCA_H is defined alloca.h will get included. now the current behaviour is to include alloca.h in any case. but if the include is protected, then the build-system has to actively define HAVE_ALLOCA_H in order keep the old behaviour. simply ignoring the other build-systems might break them. ---------------------------------------------------------------------- Comment By: Hans-Christoph Steiner (eighthave) Date: 2012-11-13 11:45 Message: I think you'll be safe only touching the ./configure.ac and Makefile.ams. Miller seems to want to hold only to the old, crufty src/configure.in build, but he doesn't use OpenBSD. ---------------------------------------------------------------------- Comment By: IOhannes m zmölnig (zmoelnig) Date: 2012-11-06 05:13 Message: #1 yes, the build-system is messed up (or rather, it is in a state of transition) #2 .../src/configure adds HAVE_foo to CPPFLAGS (and .../configure adds it to DEFS), which get expanded in the resp. makefiles (so technically you are right, they don't get added to CFLAGS) #3 i said "makefile.nt" and not "makefile.in", which is the 3rd(!) build-system that is there and which is used to build on w32 using MSVC. ---------------------------------------------------------------------- Comment By: Nobody/Anonymous (nobody) Date: 2012-11-06 04:45 Message: I might be missing something but unless you add it manually I don't see where configures adds the HAVE_XXX to CFLAGS unless you do it manually. And src/makefile.in is processed by src/configure(.in) so only ../configure.ac and ../src/configure.in need to be modified. <rant> You have a pretty messed up build system. </rant> ---------------------------------------------------------------------- Comment By: IOhannes m zmölnig (zmoelnig) Date: 2012-11-06 03:57 Message: if the patch comes from git, then please create a branch, commit to that and then create a patch-set using "git format-patch master". as for the HAVE_ALLOCA_H: if you don't instruct configure to write defines into config.h, it will add them to the CFLAGS, so that the compilation will still work. thus, no need to change the build-system. bear in mind though, that currently Pd uses to separate build-systems: .../configure.ac and .../src/configure.in (and .../src/makefile.nt); you might need to modify all 3 in order to get accepted (yes, it's a PITA) ---------------------------------------------------------------------- Comment By: Federico Schwindt (fgsch) Date: 2012-11-06 02:48 Message: @zmoelnig: The patch comes from git. My main concern is how to proceed with the other changes that require more changes than just extending an existing ifdef. For example the exclusion of alloca.h, etc. In my git clone I've modified configure to use config headers and so I can do ifdef HAVE_ALLOCA_H, etc. but I'm not sure if such patches will be accepted. The reason I've break them down is so they can be committed while I understand what your preference is. ---------------------------------------------------------------------- Comment By: IOhannes m zmölnig (zmoelnig) Date: 2012-11-06 02:39 Message: this patch is simple enough (almost trivial) as it only adds "#ifdef __OpenBSD__" to the clauses that are already handled by "__FreeBSD__". i don't see a reason to not include it (but of course this all depends on miller). small note on patches: - the patch doesn't apply here (`patch` thinks the diff is garbage) - Pd is developped in git; it would be great if you could generate your patches using git as well; see [1] - as for atomicity of patches: personally i think creating a (few) smaller patches is not a real problem; make sure that those patches are logical units rather than reflecting your iterative trial-and-error history to fix a problem. then attach all patch files to this ticket (rather than creating a new patchtracker-ticket for each file) [1] http://puredata.info/docs/developer/GitWorkflows/ ---------------------------------------------------------------------- Comment By: Federico Schwindt (fgsch) Date: 2012-11-06 02:03 Message: Comments please? I'd really to move forward with this. Thanks. ---------------------------------------------------------------------- Comment By: Federico Schwindt (fgsch) Date: 2012-10-29 19:01 Message: That'd be ideal but since I'm not sure about the other changes (using config headers, etc.) I'm submitting what I know for certain that won't change depending on the approach. For example what should I do if alloca.h doesn't exist? In my git repo I have #ifdef HAVE_ALLOCA_H where the define is driven from autotools header check but I'm still unclear whether you want to use the headers or not (and the reason you might not want to do it since you're already using autotools). Happy to submit everything in one go if you can suggest what the best course of action is. ---------------------------------------------------------------------- Comment By: Hans-Christoph Steiner (eighthave) Date: 2012-10-29 18:57 Message: The best thing to do would be to make a single patch that includes everything needed for OpenBSD support. ---------------------------------------------------------------------- You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=478072&aid=3581737&group_…

1 0

clearing out "old" build system
by Miller Puckette 20 Nov '12

20 Nov '12

To Pd developers - I'm finally gearing up to clean out the "old build system" which, I believe, will entail removing these files from pd/src: config.h.in configure configure.in install-sh makefile.clean makefile.dependencies makefile.in ... any objections? (i.e. is anyone but me still using this who can't move forward to the "new' system (cd pd; autogen.sh; ./configure; make) ? I'm planning on leaving makefile.nt and makefile.mingw around for luddites, and also will probably supply a makefile.gnu as a fallback for when, I predict inevitably, the "new build system" crumbles under its own weight. cheers Miller

3 4

[ pure-data-Patches-3581737 ] OpenBSD support
by SourceForge.net 20 Nov '12

20 Nov '12

Patches item #3581737, was opened at 2012-10-29 16:36 Message generated for change (Comment added) made by eighthave You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=478072&aid=3581737&group_… Please note that this message will contain a full copy of the comment thread, including the initial issue submission, for this request, not just the latest update. Category: puredata Group: None Status: Open Resolution: None Priority: 5 Private: No Submitted By: Federico Schwindt (fgsch) Assigned to: Nobody/Anonymous (nobody) Summary: OpenBSD support Initial Comment: FIrst round of patches to add support for OpenBSD. ---------------------------------------------------------------------- >Comment By: Hans-Christoph Steiner (eighthave) Date: 2012-11-19 19:46 Message: if you really think its worth your time, then go ahead and prepare a patch that you've tested on OpenBSD, GNU/Linux, Windows, Mac OS X, iOS, Android, Rockbox, and Microsoft Visual Studio. IMHO, things are working well as they are. ---------------------------------------------------------------------- Comment By: Federico Schwindt (fgsch) Date: 2012-11-19 15:01 Message: Not necessarily. You can supply pre-generated config.h for some platforms (that are obviously not going to change) and generate one for the rest. So basically if you're using autoconf you could use AC_CONFIG_HEADERS, if not just supply a config.h with all the defines for that platform. ---------------------------------------------------------------------- Comment By: Hans-Christoph Steiner (eighthave) Date: 2012-11-19 11:02 Message: AC_CONFIG_HEADERS works fine if your code is only built by the single autotools build system. In Pd's case, there are multiple build systems because of all of the platforms it runs on. There is configure.ac for UNIX and MinGW, src/makefile.nt for Microsoft Visual Studio, libpd for Android, iOS and more. Pd is also included in Rockbox. So using some auto-tools things makes it harder on the other platforms. ---------------------------------------------------------------------- Comment By: Miller Puckette (millerpuckette) Date: 2012-11-19 08:03 Message: I'm preparing to throw out the old "configure" stuff (which, by the way, I didn't originate and which I've tried and failed to understand; for some reason I never got AC_CONFIG_HEADERS() to work). But I'd still like to know what a good way to declare alloca(0 in a C file is. The autoconf web site suggests that I cut and past a 16-line #ifdef mess into every c file using alloca() .. perhaps under the "new" (autogen) system something shorter would work, but I'd want to have some confidence I'm doing something reasonable. ---------------------------------------------------------------------- Comment By: Federico Schwindt (fgsch) Date: 2012-11-17 19:37 Message: unfortunately you're wrong. alloca is not posix so i wouldn't expect any agreement here. as i said openbsd doesn't have alloca.h. i would not be surprised if other bsd are the same. malloc.h is also incorrect. here alloca's prototype is in stdlib.h. i will prepare a diff to add the maze of defines in all the configures and what not but I really fail to see why you can simply use AC_CONFIG_HEADERS() since you're already depending on configure. ---------------------------------------------------------------------- Comment By: Miller Puckette (millerpuckette) Date: 2012-11-17 15:03 Message: I went ahead and applied the patch by hand. I could be wrong, but isn't there now an alloca() function and an alloca.h on all unix-likes? and an alloca() defined in Windows? If so we can blow off all the configuration stuff and just conditionall include alloca or malloc.h depending on whether we're Windows. ---------------------------------------------------------------------- Comment By: IOhannes m zmölnig (zmoelnig) Date: 2012-11-13 15:45 Message: no not really. afaiu the plan is to add a check for "alloca.h" to configure, which willl define HAVE_ALLOCA_H if the file is found. if HAVE_ALLOCA_H is defined alloca.h will get included. now the current behaviour is to include alloca.h in any case. but if the include is protected, then the build-system has to actively define HAVE_ALLOCA_H in order keep the old behaviour. simply ignoring the other build-systems might break them. ---------------------------------------------------------------------- Comment By: Hans-Christoph Steiner (eighthave) Date: 2012-11-13 11:45 Message: I think you'll be safe only touching the ./configure.ac and Makefile.ams. Miller seems to want to hold only to the old, crufty src/configure.in build, but he doesn't use OpenBSD. ---------------------------------------------------------------------- Comment By: IOhannes m zmölnig (zmoelnig) Date: 2012-11-06 05:13 Message: #1 yes, the build-system is messed up (or rather, it is in a state of transition) #2 .../src/configure adds HAVE_foo to CPPFLAGS (and .../configure adds it to DEFS), which get expanded in the resp. makefiles (so technically you are right, they don't get added to CFLAGS) #3 i said "makefile.nt" and not "makefile.in", which is the 3rd(!) build-system that is there and which is used to build on w32 using MSVC. ---------------------------------------------------------------------- Comment By: Nobody/Anonymous (nobody) Date: 2012-11-06 04:45 Message: I might be missing something but unless you add it manually I don't see where configures adds the HAVE_XXX to CFLAGS unless you do it manually. And src/makefile.in is processed by src/configure(.in) so only ../configure.ac and ../src/configure.in need to be modified. <rant> You have a pretty messed up build system. </rant> ---------------------------------------------------------------------- Comment By: IOhannes m zmölnig (zmoelnig) Date: 2012-11-06 03:57 Message: if the patch comes from git, then please create a branch, commit to that and then create a patch-set using "git format-patch master". as for the HAVE_ALLOCA_H: if you don't instruct configure to write defines into config.h, it will add them to the CFLAGS, so that the compilation will still work. thus, no need to change the build-system. bear in mind though, that currently Pd uses to separate build-systems: .../configure.ac and .../src/configure.in (and .../src/makefile.nt); you might need to modify all 3 in order to get accepted (yes, it's a PITA) ---------------------------------------------------------------------- Comment By: Federico Schwindt (fgsch) Date: 2012-11-06 02:48 Message: @zmoelnig: The patch comes from git. My main concern is how to proceed with the other changes that require more changes than just extending an existing ifdef. For example the exclusion of alloca.h, etc. In my git clone I've modified configure to use config headers and so I can do ifdef HAVE_ALLOCA_H, etc. but I'm not sure if such patches will be accepted. The reason I've break them down is so they can be committed while I understand what your preference is. ---------------------------------------------------------------------- Comment By: IOhannes m zmölnig (zmoelnig) Date: 2012-11-06 02:39 Message: this patch is simple enough (almost trivial) as it only adds "#ifdef __OpenBSD__" to the clauses that are already handled by "__FreeBSD__". i don't see a reason to not include it (but of course this all depends on miller). small note on patches: - the patch doesn't apply here (`patch` thinks the diff is garbage) - Pd is developped in git; it would be great if you could generate your patches using git as well; see [1] - as for atomicity of patches: personally i think creating a (few) smaller patches is not a real problem; make sure that those patches are logical units rather than reflecting your iterative trial-and-error history to fix a problem. then attach all patch files to this ticket (rather than creating a new patchtracker-ticket for each file) [1] http://puredata.info/docs/developer/GitWorkflows/ ---------------------------------------------------------------------- Comment By: Federico Schwindt (fgsch) Date: 2012-11-06 02:03 Message: Comments please? I'd really to move forward with this. Thanks. ---------------------------------------------------------------------- Comment By: Federico Schwindt (fgsch) Date: 2012-10-29 19:01 Message: That'd be ideal but since I'm not sure about the other changes (using config headers, etc.) I'm submitting what I know for certain that won't change depending on the approach. For example what should I do if alloca.h doesn't exist? In my git repo I have #ifdef HAVE_ALLOCA_H where the define is driven from autotools header check but I'm still unclear whether you want to use the headers or not (and the reason you might not want to do it since you're already using autotools). Happy to submit everything in one go if you can suggest what the best course of action is. ---------------------------------------------------------------------- Comment By: Hans-Christoph Steiner (eighthave) Date: 2012-10-29 18:57 Message: The best thing to do would be to make a single patch that includes everything needed for OpenBSD support. ---------------------------------------------------------------------- You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=478072&aid=3581737&group_…

1 0

← Newer
1
2
3
4
5
6
7
8
9
Older →

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

2008

2007

2006

2005

2004

2003

2002

Pd-dev November 2012