I am going to take another stab at figuring out the g++
regressions in gcc trunk when built as a fink package tonight.
As I mentioned before, with the stock bison 1.28 in Tiger or
bison 2.3, I can build gcc trunk manually and it passes the
c++ testsuite fine. However so far I am finding that when fink
builds gcc trunk the c++ regressions still appear. My current
theory is that this may have something to do with how fink
sets MACOSX_DEPLOYMENT_TARGET as well as CFLAGS, CXXFLAGS
and LDFLAGS. Does anyone know what environmentals are preset
in 10.4 unstable fink during builds? I will try setting
the enviromentals to those values for a manual build of gcc
trunk and see if i can reproduce the c++ regressions that
why.
I am keen to sort this out for the following reason. If we
indeed are tickling a bug in gcc trunk by using certain settings
for the environmentals it would be wise to get a PR created for
this problem in gcc's bugzilla. If for no other reason than to
document the existance of the problem.
Jack
ps I would note that Bradley Lucier sees these regressions when
he builds gcc trunk with -mcpu=970/-m64. While I'm not using
those, I don't pass --disable-multilib so I both the 32-bit
and 64-bit libs are being built.
_______________________________________________
Fink-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/fink-devel
