Readjusted linker issues etc.

robert baker robertmbaker at gmail.com
Thu May 21 21:00:00 PDT 2009


On Wed, May 20, 2009 at 6:08 PM, Robert Connolly
<robert at linuxfromscratch.org> wrote:
>
> You should be cross referencing with lfs-svn. Sometimes they have fixes.

Good point. I will probably try to tackle that over the weekend. Up to
this point I was just trying to find the appropriate way to pass the
right options to make.

> The x86_64 additions would be nice, but it's up to you if you want to add them
> (you'll need an x86_64 to be sure it works).

Sounds interesting. I might look into that.

> I prefer 'make -C ld install'.

Then you suggest we go back to holding on to the object directory
untill the re-adjusting phase?

> Modifying the pass1 specs would use an Sed command to search for "cc1:"
> and "cc1plus:, and appending to the next new line. I don't remember how to do
> this from memory.
>
> It would be nice if upstream would add -fno-stack-protector and -fno-PIE to
> their test suite CFLAGS. They might be willing, but I don't have the first
> clue how to make a patch for it.
>
> robert

I am open to using the specs file to pass compiler flags for the test
suites. Although I think it would be just as effective to pass them to
the make command. I did some more research, and I think I found why
the gcc team isn't bothering to add those flags to their testsuite.

>From http://www.gnu.org/software/gcc/faq.html

How do I pass flags like -fnew-abi to the testsuite?
make RUNTESTFLAGS="--tool_opts '-fnew-abi -fno-honor-std'" check-g++

When I ran the tests with the below command things were considerably
better than when I used CFLAGS.

make RUNTESTFLAGS="--tool_opts '-fPIC -fno-stack-protector
-U_FORTIFY_SOURCE -nonow -nopie'" -k check



                === gcc Summary ===

# of expected passes            42538
# of unexpected failures        32  <----This was well over 300 using CFLAGS
# of expected failures          116
# of unresolved testcases       1
# of untested testcases         35
--
                === g++ Summary ===

# of expected passes            13650
# of unexpected failures        13  <----This was well over 300 as well.
# of expected failures          67
# of untested testcases         3
# of unsupported tests          77
--
                === libstdc++ Summary ===

# of expected passes            3684
# of unexpected failures        169
# of unexpected successes       1
# of expected failures          14
# of unsupported tests          316
--
                === libmudflap Summary ===

# of expected passes            1679
# of unexpected failures        106
--
                === libgomp Summary ===

# of expected passes            496

Still plenty of failures but its getting better. Did I miss any
obvious compiler options?

Robert Baker



More information about the hlfs-dev mailing list