------- Comment #5 from jay dot krell at cornell dot edu 2008-08-17 08:52
-------
Subject: RE: cannot find -lgcc_s
Subject line reminded me: Also I -disable-bootstrap.
> From: [EMAIL PROTECTED]
> To: [EMAIL PROTECTED]; [EMAIL PROTECTED]
> Subject: RE: [Bug bootstrap/37079] cannot find -lgcc_s
> Date: Sun, 17 Aug 2008 08:50:44 +0000
>
>
> Let's dissect the output a bit:
> Word wrap plus laziness => I didn't look closely before.
>
>
> -Wl,-rpath -Wl,/usr/local/lib/gcc/sparc64-sun-solaris2.10/4.3.1
> -Wl,-rpath
> -Wl,/usr/local/lib/gcc/sparc64-sun-solaris2.10/4.3.1
> -L/obj/gcc.1/sparc64-sun-solaris2.10/sparc64-sun-solaris2.10/./ld
> -L/usr/local/lib/gcc/sparc64-sun-solaris2.10/4.3.1
> -L/usr/local/lib/gcc/sparc64-sun-solaris2.10/4.3.1/../../../../sparc64-sunsolaris2.10/lib/sparcv9
> -L/usr/local/sparc64-sun-solaris2.10/sys-root/lib/sparcv9
> -L/usr/local/sparc64-sun-solaris2.10/sys-root/usr/lib/sparcv9
> -L/usr/local/lib/gcc/sparc64-sun-solaris2.10/4.3.1/../../../../sparc64-sun-solaris2.10/lib
> -L/usr/local/sparc64-sun-solaris2.10/sys-root/lib
> -L/usr/local/sparc64-sun-solaris2.10/sys-root/usr/lib
>
>
> vs. the workaround:
>
>
> Directory = Prefix + "/lib/gcc/sparc64-sun-solaris2.10/" + GccVersion
> Run(".", "mkdir -p " + Directory)
> Run(Directory, "-ln -s sparcv9/libgcc_s.so libgcc_s.so")
> Run(Directory, "-ln -s sparcv9/libgcc_s.so.1 libgcc_s.so.1")
>
>
> Therefore:
>
>
>
> The "biarch" paths are not being applied to the "gcc .libs" paths.
> /usr/local/lib/gcc/sparc64-sun-solaris2.10/4.3.1
> but not
> /usr/local/lib/gcc/sparc64-sun-solaris2.10/4.3.1/sparcv9
>
>
> both for -Wl,-rpath and -L.
> This IS reasonable, since the tools were built with -disable-multilib, BUT
> the .libs were in fact put in sparcv9 -- at least the .so files.
>
>
> (I find all the dot-dots a pain to read! I understand relocatable, and
> symlinks, but still..)
>
> They are being applied to the "system .libs" paths.
> sys-root/lib/sparcv9 and sys-root/usr/lib/sparcv9.
>
>
> Could be that .so file placement is not so flexible? Because they aren't
> really supposed to be placed here anyway? You know, it is very up to me
> where to place .a files, but .so files maybe more important to adhere
> to strict standards, since they are used at runtime?
>
>
> Also not shown is -enable-version-specific-runtimes or such, whatever
> I saw Cygwin using.
>
>
> It SEEMS the fix is now almost obvious.
> That multilib suffixes need to be applied to more prefixes.
> Or alter the directory structure otherwise, via configure.
>
>
> I should also try without -enable-rpath I guess.
>
>
> Thank you for attention to the bug,
> - Jay
>
>
>
>
>> Date: Sat, 16 Aug 2008 22:36:55 +0000
>> Subject: [Bug bootstrap/37079] cannot find -lgcc_s
>> To: [EMAIL PROTECTED]
>> From: [EMAIL PROTECTED]
>>
>>
>>
>> ------- Comment #3 from pinskia at gcc dot gnu dot org 2008-08-16 22:36
>> -------
>> What happens if you just use --with-build-sysroot= ? without saying the
>> --with-sysroot?
>>
>>
>> --
>>
>> pinskia at gcc dot gnu dot org changed:
>>
>> What |Removed |Added
>> ----------------------------------------------------------------------------
>> Component|c |bootstrap
>>
>>
>> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=37079
>>
>> ------- You are receiving this mail because: -------
>> You reported the bug, or are watching the reporter.
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=37079
