This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: Catch 5.0.x
Robert Lipe telecommunicated (on 22Dec):
|
| There are a number of chickens and eggs ranging from poached to fried
| on this topic. You've encountered only the first. The invoked
| assembler is a problem for 5.0.0 and 5.0.2 users. On 5.0.4, I think
| you're safe from them.
I build 1.0 successfully last week. Now 1.0.1 and 1.0 are both failing
if they try to use the 1.0 gcc binary. Error:
Bootstrap comparison failure!
crtbeginS.o differs
crtendS.o differs
| > Configure, in both programs, finds gcc or cc, but it then insists on
| > inserting -belf in the CFLAGS and occasionally in the LDFLAGS.
|
| Exactly at what point do you ever see it adding a -belf? When doing
| the bootstrap thing, it'll stage one with whatever you provide and then
| stage >= two with -melf becuase -melf is definitely valid in the xgcc
| that was just built.
Uh, I see it as compilation commands go by on the screen, and I see
it in the Makefiles.
| You can always override the host compiler that configure uses by specifying
| 'CC="blah options" configure'. I do this with every egcs build becuase
| configure sees that I have a gcc or two in my path and wants to use it,
| but I want to be sure everything works with the native tools. So I always
| build with 'CC=/bin/cc ../path/to/configure'.
I was just about to do that too.
| Do you have anything in your environment that could be causing autoconf
| to find -belf?
Should I suspect LD_LIBRARY_PATH=/usr/local/lib?
| > So I went around the various source files, zapped all occurences of
| > -bzip; dejanews built and installed, and make bootstrap is churning away
| > in egcs-1.0.1-gcc's stage2 even as I write.
|
| Huh? Type that again, but slower...
As I said above, it failed on a couple of file comparisons.
--
Jean-Pierre Radley jpr@jpr.com XC/XT Custodian Sysop, CompuServe SCOForum