Hi guys.
First I would like to thank everybody involved for the really nice and hard work on making this compiler rock!
Frank: you may remember, that I am a developer for Gentoo Linux (http://www.gentoo.org). Recently we announced the availablity of gcc-3.2 based profile (we skipped 3.1 series as the ABI was changed once again, though gcc guys promised to stabilize it now. We'll see :)). The gcc-2.95.x based profiles are supported still and will be supported for quite some time to come (and this is one of the niciest features of source based distro in play), so we still have the working version of gpc available.
Nonetheless, our gcc-3.2 based profile is quite stable at the moment and I would be interested in getting gpc to play nicely along. Therefore I would like to ask about the present state of the gpc development on the gcc-3 backend. I am thinking about creating the ebuild for the gcc-3 based gpc to let interested users sample and test it. The ebuild will have a "for testing" status - i.e it will not be considered stable, until you decide to make a release. We will soon have an ability to support cvs based ebuilds (that is it gets built from the present cvs code). Such ebuilds never get stable status, but are perfectly fit for this purpose. Meanwhile I can do snapshot based ebuild and periodically update it (this is generally as easy as just renaming it, provided the build procedure does not change).
Please let me know what do you think about this and where I can get the source. I tried to search the site, however I was unable to easily find the gcc-3 based source (and the 20020910 snapshot says it does not support gcc-3).
George
As an aside: As you may have read, Waldek Hebisch and I are working on the port to gcc-3 (it's not completely stable yet, but partially usable). One of the major drawbacks of gcc-2 is the internal memory management system ("obstacks") which has caused us many toubles in GPC. This bug, again, was related to obstacks, so in fact your test worked already with the gcc-3 based GPC. However, I've now fixed the real problem so it also works with obstacks (which was also necessary to fix a related problem I discovered while debugging it (fjf704.pas) and to made the code cleaner, anyway).
Frank