Re: setup 2.573.2.3 configure fails libgpg-error/libgcrypt

["Dave Korn" <dave.korn.cygwin@googlemail.com>]

Grant Edwards wrote:
> On 2009-01-13, Grant Edwards wrote:
>> I'm tyring to build setup 2.573.2.3, but the configure always
>> fails like this:
>>
>>
>>    checking whether padlock support is requested... yes
>>    checking for gpg-error-config... (cached)
>> /home/grante/setup-2.573.2.3/gpg-error-config-fake --dir=/home/grante/setup-2.573.2.3
>>    checking for GPG Error - version >= 1.4... ./configure: line 23286:
>> /home/grante/setup-2.573.2.3/gpg-error-config-fake: No such file or directory
>>    no
>>    configure: error: libgpg-error is needed.
>>                    See ftp://ftp.gnupg.org/gcrypt/libgpg-error/ .
>>    configure: error: ./configure failed for libgcrypt

  That's definitely a bit odd:

/tmp/setup-2.573.2.3/setup-2.573.2.3 $ wc -l ./configure
21895 ./configure

The error message appears to come from a couple of thousand lines after the EOF!

>> libgcrypt, libgcrypt-devel, libgcrypt11, and libgpg-error are
>> all installed.

  Yep, but as you correctly infer later, they aren't used.

>> Does anybody understand why the configure is failing?
>>
>> It appears that setup.exe includes it's own private copy of
>> libgcrypt, and it thinks there's supposed to be a file named
>> setup-2.573.2.3/gpg-error-config-fake and it can't find it?
>
> AFAICT, somebody removed the file gpg-error-config-fake
> claiming it wasn't needed anymore:
>
> http://www.cygwin.com/ml/cygwin-apps-cvs/2008-q2/msg00038.html
>
> Except that it was needed.

  That's a red herring; take a closer look at the full history:

http://cygwin.com/cgi-bin/cvsweb.cgi/setup/gpg-error-config-fake?cvsroot=cygwin-apps

  First off, that removal you cited was from mainline rather than the release
branch.  Second, it was done after applying a patch that did indeed obviate
the need for it: see r2.23 of configure.in, applied 5 minutes earlier.

http://cygwin.com/cgi-bin/cvsweb.cgi/setup/configure.in?cvsroot=cygwin-apps

  And third, I reverted both patches a few hours later after it turned out to
cause libgcrypt to get rebuilt every time you did "make" even when nothing had
changed.

  Looking in the snapshot tarball, it seems that configure.in does indeed
have the original (unpatched or reverted) code, so why doesn't it have the
gpg-error-config-fake script as well?  I went digging in the source directory
where I originally built the tarball that's currently on the cygwin website,
and the file is there in the sources, but it's not in the generated tarball!

  It turns out that adding files to the sources doesn't automatically cause
them to end up in the source tarball; you need to list them explicitly.  That
happens for you if they're C or C++ source or header files, but not for shell
scripts.  Argh, my bad!  Thanks for spotting it.

> Can somebody loan me hint on how to go about building 2.573.2.3?
>
> (or any other snapshot that is likely to work?)

  You can download a copy from the sourceware CVS repository via the web
interface here:

http://cygwin.com/cgi-bin/cvsweb.cgi/~checkout~/setup/gpg-error-config-fake?cvsroot=cygwin-apps

and drop it in with the rest of the sources.  Or you can hang around a while,
and I'll upload a fixed tarball.  (I'll just manually add the missing stuff
and re-upload it).

    cheers,
      DaveK

--
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple
Problem reports:       http://cygwin.com/problems.html
Documentation:         http://cygwin.com/docs.html
FAQ:                   http://cygwin.com/faq/