Message ID | cover.1618564630.git.fweimer@redhat.com |
---|---|
Headers | show |
Series | libpthread: Remove NPTL forwarders | expand |
* Florian Weimer via Libc-alpha: > I was on pseudo-vacation for a while and then some urgent downstream > work came up. Sorry. I hope to return to libpthread work part-time > next week, and full time after that. Hopefully this will get us back on > track. > > I believe this series addresses all of Adhemerval's review comments. It > switches to OTHER_SHLIB_COMPAT. I've retested this on i686-linux-gnu > and x86_64-linux-gnu, and built the entire thing with > build-many-glibcs.py. What's the next step for this series? Should I push it? Thanks, Florian
On Wed, Apr 21, 2021 at 6:06 AM Florian Weimer via Libc-alpha <libc-alpha@sourceware.org> wrote: > > * Florian Weimer via Libc-alpha: > > > I was on pseudo-vacation for a while and then some urgent downstream > > work came up. Sorry. I hope to return to libpthread work part-time > > next week, and full time after that. Hopefully this will get us back on > > track. > > > > I believe this series addresses all of Adhemerval's review comments. It > > switches to OTHER_SHLIB_COMPAT. I've retested this on i686-linux-gnu > > and x86_64-linux-gnu, and built the entire thing with > > build-many-glibcs.py. > > What's the next step for this series? Should I push it? > Please push it. Thanks.
The 04/21/2021 13:47, Florian Weimer via Libc-alpha wrote: > * Florian Weimer via Libc-alpha: > > > I was on pseudo-vacation for a while and then some urgent downstream > > work came up. Sorry. I hope to return to libpthread work part-time > > next week, and full time after that. Hopefully this will get us back on > > track. > > > > I believe this series addresses all of Adhemerval's review comments. It > > switches to OTHER_SHLIB_COMPAT. I've retested this on i686-linux-gnu > > and x86_64-linux-gnu, and built the entire thing with > > build-many-glibcs.py. > > What's the next step for this series? Should I push it? on arm buildbot i see build failure: a - posix/getresuid</span><span class="stderr">/tmp/ccg97Xth.s: Assembler messages: /tmp/ccg97Xth.s:210: Error: `__nptl_set_robust_list_avail' can't be equated to common symbol `__GI___nptl_set_robust_list_avail' make[2]: *** [/work/glibc-armhf-linux/build/build/sysd-rules:161: /work/glibc-armhf-linux/build/build/elf/dl-tls_init_tp.os] Error 1 </span><span class="stdout">.o
* Szabolcs Nagy: > The 04/21/2021 13:47, Florian Weimer via Libc-alpha wrote: >> * Florian Weimer via Libc-alpha: >> >> > I was on pseudo-vacation for a while and then some urgent downstream >> > work came up. Sorry. I hope to return to libpthread work part-time >> > next week, and full time after that. Hopefully this will get us back on >> > track. >> > >> > I believe this series addresses all of Adhemerval's review comments. It >> > switches to OTHER_SHLIB_COMPAT. I've retested this on i686-linux-gnu >> > and x86_64-linux-gnu, and built the entire thing with >> > build-many-glibcs.py. >> >> What's the next step for this series? Should I push it? > > on arm buildbot i see build failure: > > a - posix/getresuid</span><span class="stderr">/tmp/ccg97Xth.s: Assembler messages: > /tmp/ccg97Xth.s:210: Error: `__nptl_set_robust_list_avail' can't be equated to common symbol `__GI___nptl_set_robust_list_avail' > make[2]: *** [/work/glibc-armhf-linux/build/build/sysd-rules:161: /work/glibc-armhf-linux/build/build/elf/dl-tls_init_tp.os] Error 1 > </span><span class="stdout">.o Sorry, I didn't try building with GCC before 10 on these architectures. Please try the patch below. Thanks, Florian nptl: __nptl_set_robust_list_avail must be nocommon This is required for GCC versions before 10 which default to -fcommon. Fixes commit 442e8a40da9dfa24aeebf4f1a163f0a58b12cf7e ("nptl: Move part of TCB initialization from libpthread to __tls_init_tp"). diff --git a/sysdeps/nptl/dl-tls_init_tp.c b/sysdeps/nptl/dl-tls_init_tp.c index c5172b7613..05d2b6cfcc 100644 --- a/sysdeps/nptl/dl-tls_init_tp.c +++ b/sysdeps/nptl/dl-tls_init_tp.c @@ -23,7 +23,7 @@ #include <tls.h> #ifndef __ASSUME_SET_ROBUST_LIST -bool __nptl_set_robust_list_avail; +bool __nptl_set_robust_list_avail __attribute__ ((nocommon)); rtld_hidden_data_def (__nptl_set_robust_list_avail) #endif
The 04/22/2021 09:58, Florian Weimer wrote: > * Szabolcs Nagy: > > The 04/21/2021 13:47, Florian Weimer via Libc-alpha wrote: > >> * Florian Weimer via Libc-alpha: > >> > >> > I was on pseudo-vacation for a while and then some urgent downstream > >> > work came up. Sorry. I hope to return to libpthread work part-time > >> > next week, and full time after that. Hopefully this will get us back on > >> > track. > >> > > >> > I believe this series addresses all of Adhemerval's review comments. It > >> > switches to OTHER_SHLIB_COMPAT. I've retested this on i686-linux-gnu > >> > and x86_64-linux-gnu, and built the entire thing with > >> > build-many-glibcs.py. > >> > >> What's the next step for this series? Should I push it? > > > > on arm buildbot i see build failure: > > > > a - posix/getresuid</span><span class="stderr">/tmp/ccg97Xth.s: Assembler messages: > > /tmp/ccg97Xth.s:210: Error: `__nptl_set_robust_list_avail' can't be equated to common symbol `__GI___nptl_set_robust_list_avail' > > make[2]: *** [/work/glibc-armhf-linux/build/build/sysd-rules:161: /work/glibc-armhf-linux/build/build/elf/dl-tls_init_tp.os] Error 1 > > </span><span class="stdout">.o > > Sorry, I didn't try building with GCC before 10 on these architectures. > Please try the patch below. thanks this fixes the problem for me. > > Thanks, > Florian > > nptl: __nptl_set_robust_list_avail must be nocommon > > This is required for GCC versions before 10 which default to -fcommon. > > Fixes commit 442e8a40da9dfa24aeebf4f1a163f0a58b12cf7e ("nptl: Move part > of TCB initialization from libpthread to __tls_init_tp"). > > diff --git a/sysdeps/nptl/dl-tls_init_tp.c b/sysdeps/nptl/dl-tls_init_tp.c > index c5172b7613..05d2b6cfcc 100644 > --- a/sysdeps/nptl/dl-tls_init_tp.c > +++ b/sysdeps/nptl/dl-tls_init_tp.c > @@ -23,7 +23,7 @@ > #include <tls.h> > > #ifndef __ASSUME_SET_ROBUST_LIST > -bool __nptl_set_robust_list_avail; > +bool __nptl_set_robust_list_avail __attribute__ ((nocommon)); > rtld_hidden_data_def (__nptl_set_robust_list_avail) > #endif > --