emerge gcc:6.4.0 fails after switching to 17 profile

[sirtisimacg4]

Compiler check fails:

[Efo]

Hi sirtisimacg4,

did you use gcc-config to select gcc-6.4.0?

[Hu]

[sirtisimacg4]

It took me a while to find it (/var/tmp/portage/sys-devel/gcc-6.4.0/work/build/config.log):

localhost scott # wgetpaste config.log
Your paste can be seen here: https://paste.pound-python.org/show/S7GeBki0RtTYIxLKoI6r/

[sirtisimacg4]

Sorry for the lack of readability, using Dillo which doesn't appear to support styles - I'm trying to troubleshoot this:

These are the only errors I found in config.log

94:powerpc-unknown-linux-gnu-gcc: error: unrecognized command line option '-V'
95:powerpc-unknown-linux-gnu-gcc: fatal error: no input files
99:powerpc-unknown-linux-gnu-gcc: error: unrecognized command line option '-qversion'; did you mean '--version'?
100:powerpc-unknown-linux-gnu-gcc: fatal error: no input files
154:powerpc-unknown-linux-gnu-g++: error: unrecognized command line option '-V'
155:powerpc-unknown-linux-gnu-g++: fatal error: no input files
159:powerpc-unknown-linux-gnu-g++: error: unrecognized command line option '-qversion'; did you mean '--version'?
160:powerpc-unknown-linux-gnu-g++: fatal error: no input files

The offending lines in /var/tmp/portage/sys-devel/gcc-6.4.0/work/gcc-6.4.0/configure seem to be:

4380:for ac_option in --version -v -V -qversion; do

and

4953:for ac_option in --version -v -V -qversion; do

Build log:
wgetpaste gcc-build.log
Your paste can be seen here: https://paste.pound-python.org/show/WokNk15hZCLgoRyXnVOQ/

[image.illusion]

(This is my first time posting on the forums)

I'm having the same output(both logs and console) while trying to recompile gcc-6.4.0 in order to migrate to the 17 profile.

[sirtisimacg4]

A segmentation fault typically means you ran out of memory. Try recompiling with CFLAGS="-O1" and MAKEOPTS="-j1"

I did not see the same error in config.log. No segmentation fault in either the config.log or the gcc-build.log, but it failed in the same module as yours, so it may in fact be related. I might have gone ahead and tried changing my CFLAGS and MAKEOPTS except many posts on this forum have led me to abandon PIE altogether and implement one of the workarounds posted elsewhere.

It would not suprise me if PIE was not fully tested on such an old architecture as I'm finding this the case on more and more packages recently. The beauty of running PPC in the first place is that you are not likely going to encounter malicious PPC binaries infecting your system regardless of PIE.

[fedeliallalinea]

I'm not sure if same problem but see this bug
_________________
Questions are guaranteed in life; Answers aren't.

[image.illusion]

@sirtisimacg4 then what actually happened when the configure script invoked xgcc? The unrecognized option errors written in the log shouldn't halt the configuration.

@fedeliallalinea the circumstances are nearly identical, so I'll wait a bit for the patches. Thanks for pointing it

[sirtisimacg4]

image.illusion

It's not exactly the same as your error:

"error while loading shared libraries: R_PPC_REL24 relocation at 0xb78ba4c8 for symbol `__libc_start_main' out of range"


localhost scott # wgetpaste configlibgomp.log
Your paste can be seen here: https://paste.pound-python.org/show/4oOdmnESYWg8gf5FoiVy/

And here is the build log:
localhost scott # wgetpaste gcc-build.log
Your paste can be seen here: https://paste.pound-python.org/show/vrAXGKBCLCMANpzkRW7Q/

Yes, the config.log in .../build was not helpful


fedeliallalinea: Thanks for the info on the bug. This looks like a show stopper for ppc and PIE until patched