contaminated build environment for gtk2 (was Re: error building gtk2 on Mac OS 10.6.4)

["Mark E. Perkins" <perkinsm%bway.net@localhost>]

On 2010/11/28 09:40, Mark E. Perkins wrote:
> On 2010/11/28 08:49, Mark E. Perkins wrote:
>> I have pkgsrc-2010Q3 refreshed earlier today.  Building x11/gtk2, I get this:
>>
>> ===> Configuring for gtk2+-2.20.1nb2
>> => Fixing pthread flags.
>> => Modifying GNU configure scripts to avoid --recheck
>> => Replacing config-guess with pkgsrc versions
>> => Replacing config-sub with pkgsrc versions
>> => Replacing install-sh with pkgsrc version
>> => Replacing python interpreter in gtk/gtk-builder-convert.
>> => Fixing locale directory references.
>> configure: WARNING: unrecognized options: --with-libintl-prefix,
>> --with-libiconv-prefix
>> configure: error: pwd does not report name of working directory
>> *** Error code 1
>>
>> Stop.
>> bmake: stopped in /usr/pkgsrc/x11/gtk2
>> *** Error code 1
>>
>>
>> work/.tools/bin/pwd points to /bin/pwd which works correctly AFAICT.  It
>> looks to me that the tests in configure should pass, but the last
>> comparison ( test ... || as_fnerror ... ) fails and I can't see why.
>>
>> Any suggestions for a fix (or where to look for more info)?
>>
>> Thanks,
>> Mark
> 
> I also see that the same tests pass during configure of other packages.

The problem here turns out to be leakage of settings from my usual working
environment into the build enviroment for gtk2.  In brief, I have a (bash)
function 'cd' to fiddle with xterm window titles and 'cd' is used in the
(failing) test of whether 'pwd' works properly.  Solve that by commenting
out the relevant lines in ~/.bashrc (not sufficient to 'unset -f cd', so it
appears that my .bashrc is sourced somewhere along the way).  That gets me
past the 'pwd' check OK, but then 'configure' fails a check on 'ls' (I have
that aliased to 'ls -F').  Again, comment out of ~/.bashrc, and then build
of gtk2 runs successfully.

My recollection is that pkgsrc scrubs the environment when invoked, but I'm
not familiar with the lower level details.  As noted above, those same
tests are included in other configure scripts and they don't result in
failures, suggesting that something is amiss in the way building gtk2
cleans (or not) before it gets going.

I will add a few details to above comments and submit as a PR so this can
be tracked.

Cheers,
Mark