diff mbox series

[2/2] Avoid adding duplicated symbols into static libraries

Message ID 20210211101908.2825685-3-siddhesh@sourceware.org
State Committed
Commit 4898d9712bbd85e6fb576442f578d6f3c3e35898
Headers
Series math symbols cleanup |

Commit Message

Siddhesh Poyarekar Feb. 11, 2021, 10:19 a.m. UTC 
  Some math functions (such as __isnan*) are built into both libm and
libc because they are needed in libc.  The symbol gets exported from
libc.so and not libm.so, because of which dynamic linking works fine;
the symbols are always resolved from libc.so and libm.so uses its
internal copy of the same function if needed.

When linking statically though, the libm variants get used throughout
because the symbols are exported in both archives and libm.a is
searched first.

This patch removes these duplicate objects from the libm.a archive so
that programs always link to libc in both, the static and dynamic
case.  The difference this will cause is that libm uses of these
functions will start using the libc versions in the !SHARED case.
This is harmless at the moment because the objects are identical
except for their names.

Some of these duplicates could be removed from libm.so too, but I
avoided that in the interest of retaining an internal reference if at
all those functions get used within libm in future.
---
 math/Makefile | 3 +++
 1 file changed, 3 insertions(+)

Comments

Adhemerval Zanella March 30, 2021, 7:47 p.m. UTC | #1 
On 11/02/2021 07:19, Siddhesh Poyarekar via Libc-alpha wrote:
> Some math functions (such as __isnan*) are built into both libm and
> libc because they are needed in libc.  The symbol gets exported from
> libc.so and not libm.so, because of which dynamic linking works fine;
> the symbols are always resolved from libc.so and libm.so uses its
> internal copy of the same function if needed.
> 
> When linking statically though, the libm variants get used throughout
> because the symbols are exported in both archives and libm.a is
> searched first.
> 
> This patch removes these duplicate objects from the libm.a archive so
> that programs always link to libc in both, the static and dynamic
> case.  The difference this will cause is that libm uses of these
> functions will start using the libc versions in the !SHARED case.
> This is harmless at the moment because the objects are identical
> except for their names.
> 
> Some of these duplicates could be removed from libm.so too, but I
> avoided that in the interest of retaining an internal reference if at
> all those functions get used within libm in future.
> ---
>  math/Makefile | 3 +++
>  1 file changed, 3 insertions(+)
> 
> diff --git a/math/Makefile b/math/Makefile
> index 687aa5d510..c4cc1b2068 100644
> --- a/math/Makefile
> +++ b/math/Makefile
> @@ -196,6 +196,9 @@ calls = s_isinfF s_isnanF s_finiteF s_copysignF s_modfF s_scalbnF s_frexpF \
>  gen-calls = s_ldexpF
>  generated += $(foreach s,.c .S,$(call type-foreach, $(calls:s_%=m_%$(s))))
>  routines = $(call type-foreach, $(calls))
> +# The $(calls) that are shared between libm and libc are not included in static
> +# libm so the symbols end up in exactly one place.
> +libm-shared-only-routines = $(call type-foreach, $(calls:s_%=m_%))
>  
>  ifeq ($(build-mathvec),yes)
>  # We need to install libm.so and libm.a as linker scripts
> 

This broke powerpc64le static math tests:

powerpc64le-linux-gnu]$ make math/tests
[...]
/home/azanella/toolchain/install/compilers/powerpc64le-linux-gnu/bin/powerpc64le-glibc-linux-gnu-gcc -nostdlib -nostartfiles -static -o /home/azanella/glibc/build/powerpc64le-linux-gnu/math/test-signgam-uchar-static     /home/azanella/glibc/build/powerpc64le-linux-gnu/csu/crt1.o /home/azanella/glibc/build/powerpc64le-linux-gnu/csu/crti.o `/home/azanella/toolchain/install/compilers/powerpc64le-linux-gnu/bin/powerpc64le-glibc-linux-gnu-gcc  --print-file-name=crtbeginT.o` /home/azanella/glibc/build/powerpc64le-linux-gnu/math/test-signgam-uchar-static.o /home/azanella/glibc/build/powerpc64le-linux-gnu/support/libsupport_nonshared.a /home/azanella/glibc/build/powerpc64le-linux-gnu/math/libm.a  -Wl,--start-group /home/azanella/glibc/build/powerpc64le-linux-gnu/libc.a -lgcc -lgcc_eh  -Wl,--end-group `/home/azanella/toolchain/install/compilers/powerpc64le-linux-gnu/bin/powerpc64le-glibc-linux-gnu-gcc  --print-file-name=crtend.o` /home/azanella/glibc/build/powerpc64le-linux-gnu/csu/crtn.o
/home/azanella/toolchain/install/compilers/powerpc64le-linux-gnu/lib/gcc/powerpc64le-glibc-linux-gnu/10.2.1/../../../../powerpc64le-glibc-linux-gnu/bin/ld: /home/azanella/glibc/build/powerpc64le-linux-gnu/math/libm.a(e_logf128.o): in function `__ieee754_logf128_power8':
/home/azanella/glibc/glibc-git/math/../sysdeps/ieee754/float128/../ldbl-128/e_logl.c:217: undefined reference to `__frexpf128_power8'
/home/azanella/toolchain/install/compilers/powerpc64le-linux-gnu/lib/gcc/powerpc64le-glibc-linux-gnu/10.2.1/../../../../powerpc64le-glibc-linux-gnu/bin/ld: /home/azanella/glibc/build/powerpc64le-linux-gnu/math/libm.a(s_log1pf128.o): in function `__log1pf128_power8':
/home/azanella/glibc/glibc-git/math/../sysdeps/ieee754/float128/../ldbl-128/s_log1pl.c:166: undefined reference to `__frexpf128_power8'
collect2: error: ld returned 1 exit status
make[2]: *** [../Rules:264: /home/azanella/glibc/build/powerpc64le-linux-gnu/math/test-signgam-uchar-static] Error 1
Siddhesh Poyarekar March 31, 2021, 2:57 a.m. UTC | #2 
On 3/31/21 1:17 AM, Adhemerval Zanella wrote:
> This broke powerpc64le static math tests:
> 
> powerpc64le-linux-gnu]$ make math/tests
> [...]
> /home/azanella/toolchain/install/compilers/powerpc64le-linux-gnu/bin/powerpc64le-glibc-linux-gnu-gcc -nostdlib -nostartfiles -static -o /home/azanella/glibc/build/powerpc64le-linux-gnu/math/test-signgam-uchar-static     /home/azanella/glibc/build/powerpc64le-linux-gnu/csu/crt1.o /home/azanella/glibc/build/powerpc64le-linux-gnu/csu/crti.o `/home/azanella/toolchain/install/compilers/powerpc64le-linux-gnu/bin/powerpc64le-glibc-linux-gnu-gcc  --print-file-name=crtbeginT.o` /home/azanella/glibc/build/powerpc64le-linux-gnu/math/test-signgam-uchar-static.o /home/azanella/glibc/build/powerpc64le-linux-gnu/support/libsupport_nonshared.a /home/azanella/glibc/build/powerpc64le-linux-gnu/math/libm.a  -Wl,--start-group /home/azanella/glibc/build/powerpc64le-linux-gnu/libc.a -lgcc -lgcc_eh  -Wl,--end-group `/home/azanella/toolchain/install/compilers/powerpc64le-linux-gnu/bin/powerpc64le-glibc-linux-gnu-gcc  --print-file-name=crtend.o` /home/azanella/glibc/build/powerpc64le-linux-gnu/csu/crtn.o
> /home/azanella/toolchain/install/compilers/powerpc64le-linux-gnu/lib/gcc/powerpc64le-glibc-linux-gnu/10.2.1/../../../../powerpc64le-glibc-linux-gnu/bin/ld: /home/azanella/glibc/build/powerpc64le-linux-gnu/math/libm.a(e_logf128.o): in function `__ieee754_logf128_power8':
> /home/azanella/glibc/glibc-git/math/../sysdeps/ieee754/float128/../ldbl-128/e_logl.c:217: undefined reference to `__frexpf128_power8'
> /home/azanella/toolchain/install/compilers/powerpc64le-linux-gnu/lib/gcc/powerpc64le-glibc-linux-gnu/10.2.1/../../../../powerpc64le-glibc-linux-gnu/bin/ld: /home/azanella/glibc/build/powerpc64le-linux-gnu/math/libm.a(s_log1pf128.o): in function `__log1pf128_power8':
> /home/azanella/glibc/glibc-git/math/../sysdeps/ieee754/float128/../ldbl-128/s_log1pl.c:166: undefined reference to `__frexpf128_power8'
> collect2: error: ld returned 1 exit status
> make[2]: *** [../Rules:264: /home/azanella/glibc/build/powerpc64le-linux-gnu/math/test-signgam-uchar-static] Error 1

Sorry about that.  I'll revert the patch if I'm unable to post a fix by 
the end of the day.

Siddhesh
Tulio Magno Quites Machado Filho March 31, 2021, 1:54 p.m. UTC | #3 
Siddhesh Poyarekar via Libc-alpha <libc-alpha@sourceware.org> writes:

> On 3/31/21 1:17 AM, Adhemerval Zanella wrote:
>> This broke powerpc64le static math tests:
>> 
>> powerpc64le-linux-gnu]$ make math/tests
>> [...]
>> /home/azanella/toolchain/install/compilers/powerpc64le-linux-gnu/bin/powerpc64le-glibc-linux-gnu-gcc -nostdlib -nostartfiles -static -o /home/azanella/glibc/build/powerpc64le-linux-gnu/math/test-signgam-uchar-static     /home/azanella/glibc/build/powerpc64le-linux-gnu/csu/crt1.o /home/azanella/glibc/build/powerpc64le-linux-gnu/csu/crti.o `/home/azanella/toolchain/install/compilers/powerpc64le-linux-gnu/bin/powerpc64le-glibc-linux-gnu-gcc  --print-file-name=crtbeginT.o` /home/azanella/glibc/build/powerpc64le-linux-gnu/math/test-signgam-uchar-static.o /home/azanella/glibc/build/powerpc64le-linux-gnu/support/libsupport_nonshared.a /home/azanella/glibc/build/powerpc64le-linux-gnu/math/libm.a  -Wl,--start-group /home/azanella/glibc/build/powerpc64le-linux-gnu/libc.a -lgcc -lgcc_eh  -Wl,--end-group `/home/azanella/toolchain/install/compilers/powerpc64le-linux-gnu/bin/powerpc64le-glibc-linux-gnu-gcc  --print-file-name=crtend.o` /home/azanella/glibc/build/powerpc64le-linux-gnu/csu/crtn.o
>> /home/azanella/toolchain/install/compilers/powerpc64le-linux-gnu/lib/gcc/powerpc64le-glibc-linux-gnu/10.2.1/../../../../powerpc64le-glibc-linux-gnu/bin/ld: /home/azanella/glibc/build/powerpc64le-linux-gnu/math/libm.a(e_logf128.o): in function `__ieee754_logf128_power8':
>> /home/azanella/glibc/glibc-git/math/../sysdeps/ieee754/float128/../ldbl-128/e_logl.c:217: undefined reference to `__frexpf128_power8'
>> /home/azanella/toolchain/install/compilers/powerpc64le-linux-gnu/lib/gcc/powerpc64le-glibc-linux-gnu/10.2.1/../../../../powerpc64le-glibc-linux-gnu/bin/ld: /home/azanella/glibc/build/powerpc64le-linux-gnu/math/libm.a(s_log1pf128.o): in function `__log1pf128_power8':
>> /home/azanella/glibc/glibc-git/math/../sysdeps/ieee754/float128/../ldbl-128/s_log1pl.c:166: undefined reference to `__frexpf128_power8'
>> collect2: error: ld returned 1 exit status
>> make[2]: *** [../Rules:264: /home/azanella/glibc/build/powerpc64le-linux-gnu/math/test-signgam-uchar-static] Error 1
>
> Sorry about that.  I'll revert the patch if I'm unable to post a fix by 
> the end of the day.

I don't think we need to hurry and revert this.
I'm also looking at this.  It has to do with multiarch builds.

I'm sorry for not taking a look at patch 2/2.  :-(
Siddhesh Poyarekar March 31, 2021, 2:05 p.m. UTC | #4 
On 3/31/21 7:24 PM, Tulio Magno Quites Machado Filho via Libc-alpha wrote:
> I don't think we need to hurry and revert this.
> I'm also looking at this.  It has to do with multiarch builds.
> 
> I'm sorry for not taking a look at patch 2/2.  :-(

Sorry I rushed to push this :)

I posted a fix to patch this up but I think Andreas may have a more 
complete fix.  Could you please review that one?

https://sourceware.org/pipermail/libc-alpha/2021-March/124602.html

Thanks,
Siddhesh
diff mbox series

Patch

diff --git a/math/Makefile b/math/Makefile
index 687aa5d510..c4cc1b2068 100644
--- a/math/Makefile
+++ b/math/Makefile
@@ -196,6 +196,9 @@  calls = s_isinfF s_isnanF s_finiteF s_copysignF s_modfF s_scalbnF s_frexpF \
 gen-calls = s_ldexpF
 generated += $(foreach s,.c .S,$(call type-foreach, $(calls:s_%=m_%$(s))))
 routines = $(call type-foreach, $(calls))
+# The $(calls) that are shared between libm and libc are not included in static
+# libm so the symbols end up in exactly one place.
+libm-shared-only-routines = $(call type-foreach, $(calls:s_%=m_%))
 
 ifeq ($(build-mathvec),yes)
 # We need to install libm.so and libm.a as linker scripts