I'm CC'ing pd-dev since this is good to have in the public and on the record.

On Dec 7, 2012, at 3:51 AM, IOhannes m zmölnig wrote:

On 12/07/2012 09:22, IOhannes m zmölnig wrote:

...

hi hans!

...

do you have any idea what causes such "spurious" build failures on the

jenkins machines?

a related issue, is the build-failure of puredata on the macosx104-powerpc machine, which is consistent and reproducible.

i checked and the problem is, that some pkg-config related m4 macros are not expanded correctly.

the relevant m4 macro get's installed by fink's "pkg-config" package as /sw/share/aclocal/pkg.m4

unfortunately, the "aclocal" binary used on the machine is "/usr/bin/aclocal" (that is: not-fink), and does not automatically look into "/sw/share/aclocal" for additional m4 macros.

i don't know what the proper way to deal with this is:

• either install a pkg-config version of "aclocal" (which will automatically search the right paths))
• or tell the system's aclocal to use fink's m4-directory (adding "-I /sw/share/aclocal" to the autoreconf/aclocal flags)

i don't like the latter much, as i assume it will be hard to get it right (e.g. not using m4 macros in /sw when the user decides to *not* activate fink even though it's installed on the machine), but i don't know whether the first is actually possible.

I think the problem was that Fink's 'autoconf' was installed but not Fink's 'automake-1.10'. So we need to decide to either go full native or full Fink for the autotools. The minimum build env I support is 10.5, so that means:

autoconf 2.61 automake 1.10 libtool 1.5.22

If things need newer versions than that, then I say we just use all of the Fink packages. Then we could have:

autoconf 2.69 automake 1.12.3 libtool 2.4.2

.hc