On Fri, Dec 02, 2011 at 09:46:22PM +0000, corvid wrote:
Jorge wrote:
AFAIR, libpng detection was an issue (not only for us), and it took a long time to come with something that worked across all the different platforms we were testing.
What we can glean from good old CVS is:
In December 2004, the test changed from which libpng12-config || which libpng-config || which libpng10-config to a bunch of separate whiches because "Fixed a libpng detection problem (e.g., on CYGWIN). BUG#651". Of course, BUG#651 is long gone, but presumably the problem was the "||".
In January 2004, we go from PNG_CONFIG="`which libpng-config || which libpng12-config`" to PNG_CONFIG=`which libpng12-config || which libpng-config || which libpng10-config` (along with general restructuring) Then if we follow that to the archive, http://lists.auriga.wearlab.de/pipermail/dillo-dev/2004-January/001683.html the libpng talk is of Slackware 3.9, which came out in May, 1999. A quoted msg from Jorgen doesn't appear to be in the archive.
Looking at http://slackware.cs.utah.edu/pub/slackware/slackware-3.9/slakware/MANIFEST.g... Slackware 3.9 used libpng-1.0.3, and did not have a png-config of any sort that I can see.
Looking at the CHANGES file with my copy of libpng shows that png-config showed up during the 1.2.x series.
Earlier in January, http://lists.auriga.wearlab.de/pipermail/dillo-dev/2004-January/001615.html says that libpng 1.2.5 needed to be added to the Slackware 3.9 system to make dillo happy.
Overall, the messages give every appearance of saying that dillo should start with trying the libpng-config system but then still work on systems that predate libpng-config.
Looking at the change made, yeah, I guess that pre-libpng-config system mustn't have had libpng stuff in the right places to get by with CPPFLAGS as they were, hence added code that, if there was no libpng-config at all and there was png.h, then add in 'LIBPNG_LIBS="-lpng -lm"' ourselves.
libpng configuration checking was introduced in March 2003. This leads back to http://lists.auriga.wearlab.de/pipermail/dillo-dev/2003-March/000390.html where nothing brought up seems to have bearing on the matter at hand here.
Nothing reveals itself in the extant archaeological record suggesting that the various libpng-configs themselves were an issue.
Sorry if my meaning/wording wasn't clear. I meant, "it took some effort to sort the detection code so it finally was able to compile automatically accross all the different platforms we were testing" (some had libpng-config, some not, some had both installed at the same time IIRC). -- Cheers Jorge.-