autoconf / automake: apply patches by KOMH

[dmik]

There are fresh versions of autoconf and automake from KOMH:

​http://hobbes.nmsu.edu/download/pub/os2/dev/util/automake-1.13.1.zip
​http://hobbes.nmsu.edu/download/pub/os2/dev/util/autoconf-2.69.zip

We should update our repos and RPMs.

[dmik]

From what I see, all OS/2 patches from autoconf-2.69.zip are already applied in r218. So I only need to update SVN to v2.69. automake-1.13.1.zip seems to contain no OS/2 patches at all but I made some OS/2-specific fixes there that are needed to make it work. These fixes require autoconf 2.69 though so I will do that first.

[dmik]

I can't properly build autoconf here since there are tons of various problems. E.g. it seems that the m4_esyscmd macro doesn't work properly and as a result it's impossible to run the make _version target that is supposed to update the .version file with the current version.

Also, there seems to be many more broken things. At least, neither autoconf 2.69 nor autoconf 2.65 are able to properly generate a working configure for expat. The resulting configure complains that the LIBTOOL is missing and such.

[dmik]

Trying to rebuild m4 in order to debug the m4_esyscmd macro and it of course fails:

checking whether an open file can be renamed... return: Illegal number: -/

I'm very close to giving up.

[dmik]

Okay, GREAT. There is a bug in RPM SH.EXE: If an EXE program returns a negative return code, SH sets the value of $? to garbage (-/ in this case). I will create a separate ticket for SH.

So, my current work's dependency is as follows (assuming that there are no new perl bugs):

git -> expat -> automake -> autoconf -> m4 -> sh.

Okay. But this will take way more than we initially planned for git of course (mid of June).

[dmik]

Created #27 for the SH bug.

[dmik]

The fun thing is that the RPM build of m4 doesn't fail because it uses the urpo library from Yuri Dario that allows to rename open files and this makes the test checking whether an open file can be renamed... to succeed (and return 0 instead of a negative number) which in turn does not cause the SH bug (since positive exit codes are handled well).

[dmik]

Seems that the RPM build of m4 is somehow broken. Created #28 for that.

[dmik]

Added automake to SVN with my own fixes required to run it in an RPM env on OS/2 . Also improved several things in autoconf (see the timeline). Will create new RPM packages in a couple of days.