On 08/23/13 21:00, Miller Puckette wrote:
Hi all,
Pd version 0.45-0 is available on http://crca.ucsd.edu/~msp/software.htm
cool. i started updating the Debian packages...
As always I'm sure there will be problems here and there - you're welcome to report them on the Pd mailing list (pd-list@iem.at) which is always the fastest way to get me to see them.
two smallish notes.
#1 CFLAGS when manually including the Debian-patch "usercflags" (that makes sure that user-specified CFLAGS are honoured by prepending the configure-detected flags rather than appending), a small oversight has occured:
- on *linux*, the configure-CFLAGS are still appended. instead the
change was made for *hurd* (aka "GNU"). i think this was simply a confusion of "GNU" and "linux".
- otoh. the *hurd* CFLAGS still include "-O6"
the attached "usercflags.patch" hopefully gets this right. it /also/ fixes the CFLAGS append/prepend order for cygwin & mingw (though i admit that i have not tested this)
#2 BUILD FAILURES with "-Werror=format-security" when trying to build with the above error-flag (which pays some extra attention to argument passing), the build fails due to two problems. a) a call of "error(buf)" withing bug() is rejected, as the 'error'-function really reads "error(const char *fmt, ...)" and 'buf' is not a format, resp. there are no varargs. the fix is simply to call error("consistency check failed: %s", buf) and not construct the "consitency-failed" string beforehand. cool, this makes the code a little more readable! b) more serious, the new [text] object uses 'pd_error("ouch %s", str)', when pd_error() really needs a pointer to a pd-object... this is a potential crasher bug, as it accesses string memory as objects.
the attached patch "fix_format-security.patch" fixes this as well.
gfmasdr IOhannes