* [PATCH] resolv: Do not build libanl.so for ABIs starting at 2.35 @ 2021-12-29 16:39 Adhemerval Zanella 2021-12-29 18:30 ` Florian Weimer 0 siblings, 1 reply; 7+ messages in thread From: Adhemerval Zanella @ 2021-12-29 16:39 UTC (permalink / raw) To: libc-alpha, Florian Weimer --- resolv/Makefile | 11 ++++++++++- 1 file changed, 10 insertions(+), 1 deletion(-) diff --git a/resolv/Makefile b/resolv/Makefile index 59e599535c..5606eab1fd 100644 --- a/resolv/Makefile +++ b/resolv/Makefile @@ -78,8 +78,12 @@ generate := mtrace-tst-leaks.out tst-leaks.mtrace tst-leaks2.mtrace extra-libs := libresolv libnss_dns ifeq ($(have-thread-library),yes) -extra-libs += libanl routines += gai_sigqueue +endif + +ifeq ($(have-GLIBC_2.34)$(have-thread-library),yesyes) +# Empty compatibility library for old binaries. +extra-libs += libanl tests += \ tst-bug18665 \ @@ -176,6 +180,11 @@ $(libanl-routines-var) += \ libanl-routines += libanl-compat libanl-shared-only-routines += libanl-compat +# Pretend that libanl.so is a linker script, so that the symbolic link +# is not installed. +install-lib-ldscripts = libanl.so +$(inst_libdir)/libanl.so: + subdir-dirs = nss_dns vpath %.c nss_dns -- 2.32.0 ^ permalink raw reply [flat|nested] 7+ messages in thread
* Re: [PATCH] resolv: Do not build libanl.so for ABIs starting at 2.35 2021-12-29 16:39 [PATCH] resolv: Do not build libanl.so for ABIs starting at 2.35 Adhemerval Zanella @ 2021-12-29 18:30 ` Florian Weimer 2021-12-30 14:06 ` Adhemerval Zanella 0 siblings, 1 reply; 7+ messages in thread From: Florian Weimer @ 2021-12-29 18:30 UTC (permalink / raw) To: Adhemerval Zanella; +Cc: libc-alpha * Adhemerval Zanella: > --- > resolv/Makefile | 11 ++++++++++- > 1 file changed, 10 insertions(+), 1 deletion(-) > > diff --git a/resolv/Makefile b/resolv/Makefile > index 59e599535c..5606eab1fd 100644 > --- a/resolv/Makefile > +++ b/resolv/Makefile > @@ -78,8 +78,12 @@ generate := mtrace-tst-leaks.out tst-leaks.mtrace tst-leaks2.mtrace > > extra-libs := libresolv libnss_dns > ifeq ($(have-thread-library),yes) > -extra-libs += libanl > routines += gai_sigqueue > +endif > + > +ifeq ($(have-GLIBC_2.34)$(have-thread-library),yesyes) > +# Empty compatibility library for old binaries. > +extra-libs += libanl > > tests += \ > tst-bug18665 \ > @@ -176,6 +180,11 @@ $(libanl-routines-var) += \ > libanl-routines += libanl-compat > libanl-shared-only-routines += libanl-compat > > +# Pretend that libanl.so is a linker script, so that the symbolic link > +# is not installed. > +install-lib-ldscripts = libanl.so > +$(inst_libdir)/libanl.so: > + Looks okay. But the second hunk should be in a separate commit; I think we should backport it. Thanks, Florian ^ permalink raw reply [flat|nested] 7+ messages in thread
* Re: [PATCH] resolv: Do not build libanl.so for ABIs starting at 2.35 2021-12-29 18:30 ` Florian Weimer @ 2021-12-30 14:06 ` Adhemerval Zanella 2022-01-03 4:53 ` Stafford Horne 0 siblings, 1 reply; 7+ messages in thread From: Adhemerval Zanella @ 2021-12-30 14:06 UTC (permalink / raw) To: Florian Weimer; +Cc: libc-alpha On 29/12/2021 15:30, Florian Weimer wrote: > * Adhemerval Zanella: > >> --- >> resolv/Makefile | 11 ++++++++++- >> 1 file changed, 10 insertions(+), 1 deletion(-) >> >> diff --git a/resolv/Makefile b/resolv/Makefile >> index 59e599535c..5606eab1fd 100644 >> --- a/resolv/Makefile >> +++ b/resolv/Makefile >> @@ -78,8 +78,12 @@ generate := mtrace-tst-leaks.out tst-leaks.mtrace tst-leaks2.mtrace >> >> extra-libs := libresolv libnss_dns >> ifeq ($(have-thread-library),yes) >> -extra-libs += libanl >> routines += gai_sigqueue >> +endif >> + >> +ifeq ($(have-GLIBC_2.34)$(have-thread-library),yesyes) >> +# Empty compatibility library for old binaries. >> +extra-libs += libanl >> >> tests += \ >> tst-bug18665 \ >> @@ -176,6 +180,11 @@ $(libanl-routines-var) += \ >> libanl-routines += libanl-compat >> libanl-shared-only-routines += libanl-compat >> >> +# Pretend that libanl.so is a linker script, so that the symbolic link >> +# is not installed. >> +install-lib-ldscripts = libanl.so >> +$(inst_libdir)/libanl.so: >> + > > Looks okay. But the second hunk should be in a separate commit; I think > we should backport it. Done. ^ permalink raw reply [flat|nested] 7+ messages in thread
* Re: [PATCH] resolv: Do not build libanl.so for ABIs starting at 2.35 2021-12-30 14:06 ` Adhemerval Zanella @ 2022-01-03 4:53 ` Stafford Horne 2022-01-03 13:58 ` Adhemerval Zanella 0 siblings, 1 reply; 7+ messages in thread From: Stafford Horne @ 2022-01-03 4:53 UTC (permalink / raw) To: Adhemerval Zanella; +Cc: Florian Weimer, libc-alpha On Thu, Dec 30, 2021 at 11:06:31AM -0300, Adhemerval Zanella via Libc-alpha wrote: > > > On 29/12/2021 15:30, Florian Weimer wrote: > > * Adhemerval Zanella: > > > >> --- > >> resolv/Makefile | 11 ++++++++++- > >> 1 file changed, 10 insertions(+), 1 deletion(-) > >> > >> diff --git a/resolv/Makefile b/resolv/Makefile > >> index 59e599535c..5606eab1fd 100644 > >> --- a/resolv/Makefile > >> +++ b/resolv/Makefile > >> @@ -78,8 +78,12 @@ generate := mtrace-tst-leaks.out tst-leaks.mtrace tst-leaks2.mtrace > >> > >> extra-libs := libresolv libnss_dns > >> ifeq ($(have-thread-library),yes) > >> -extra-libs += libanl > >> routines += gai_sigqueue > >> +endif > >> + > >> +ifeq ($(have-GLIBC_2.34)$(have-thread-library),yesyes) > >> +# Empty compatibility library for old binaries. > >> +extra-libs += libanl > >> > >> tests += \ > >> tst-bug18665 \ > >> @@ -176,6 +180,11 @@ $(libanl-routines-var) += \ > >> libanl-routines += libanl-compat > >> libanl-shared-only-routines += libanl-compat > >> > >> +# Pretend that libanl.so is a linker script, so that the symbolic link > >> +# is not installed. > >> +install-lib-ldscripts = libanl.so > >> +$(inst_libdir)/libanl.so: > >> + > > > > Looks okay. But the second hunk should be in a separate commit; I think > > we should backport it. > > Done. After this make check fails on the OpenRISC port. tst-linkall-static fails with: make[3]: *** No rule to make target 'build/glibcs/or1k-linux-gnu-soft/glibc/resolv/libanl.a', needed by 'build/glibcs/or1k-linux-gnu-soft/glibc/elf/tst-linkall-static'. Stop. We may want to keep libanl.a around as an empty binary for build compatibility reasons like we did with libutil. There may be builds wheck hard code -lanl, maybe not as much as -lutil though. Never the less, this patch seems to fix it: diff --git a/elf/Makefile b/elf/Makefile index 06bfa1642f..71bc2987df 100644 --- a/elf/Makefile +++ b/elf/Makefile @@ -1683,7 +1683,6 @@ $(objpfx)tst-linkall-static: \ $(common-objpfx)resolv/libresolv.a \ $(common-objpfx)login/libutil.a \ $(common-objpfx)rt/librt.a \ - $(common-objpfx)resolv/libanl.a \ $(static-thread-library) ifeq ($(build-crypt),yes) @@ -1704,6 +1703,12 @@ $(objpfx)tst-linkall-static: \ endif endif +# Libanl is only available up to GLIBC_2.34 +ifeq ($(have-GLIBC_2.34)$(have-thread-library),yesyes) +$(objpfx)tst-linkall-static: \ + $(common-objpfx)resolv/libanl.a +endif + # The application depends on the DSO, and the DSO loads the plugin. # The plugin also depends on the DSO. This creates the circular # dependency via dlopen that we're testing to make sure works. ^ permalink raw reply [flat|nested] 7+ messages in thread
* Re: [PATCH] resolv: Do not build libanl.so for ABIs starting at 2.35 2022-01-03 4:53 ` Stafford Horne @ 2022-01-03 13:58 ` Adhemerval Zanella 2022-02-11 11:34 ` Stafford Horne 0 siblings, 1 reply; 7+ messages in thread From: Adhemerval Zanella @ 2022-01-03 13:58 UTC (permalink / raw) To: Stafford Horne; +Cc: Florian Weimer, libc-alpha On 03/01/2022 01:53, Stafford Horne wrote: > On Thu, Dec 30, 2021 at 11:06:31AM -0300, Adhemerval Zanella via Libc-alpha wrote: >> >> >> On 29/12/2021 15:30, Florian Weimer wrote: >>> * Adhemerval Zanella: >>> >>>> --- >>>> resolv/Makefile | 11 ++++++++++- >>>> 1 file changed, 10 insertions(+), 1 deletion(-) >>>> >>>> diff --git a/resolv/Makefile b/resolv/Makefile >>>> index 59e599535c..5606eab1fd 100644 >>>> --- a/resolv/Makefile >>>> +++ b/resolv/Makefile >>>> @@ -78,8 +78,12 @@ generate := mtrace-tst-leaks.out tst-leaks.mtrace tst-leaks2.mtrace >>>> >>>> extra-libs := libresolv libnss_dns >>>> ifeq ($(have-thread-library),yes) >>>> -extra-libs += libanl >>>> routines += gai_sigqueue >>>> +endif >>>> + >>>> +ifeq ($(have-GLIBC_2.34)$(have-thread-library),yesyes) >>>> +# Empty compatibility library for old binaries. >>>> +extra-libs += libanl >>>> >>>> tests += \ >>>> tst-bug18665 \ >>>> @@ -176,6 +180,11 @@ $(libanl-routines-var) += \ >>>> libanl-routines += libanl-compat >>>> libanl-shared-only-routines += libanl-compat >>>> >>>> +# Pretend that libanl.so is a linker script, so that the symbolic link >>>> +# is not installed. >>>> +install-lib-ldscripts = libanl.so >>>> +$(inst_libdir)/libanl.so: >>>> + >>> >>> Looks okay. But the second hunk should be in a separate commit; I think >>> we should backport it. >> >> Done. > > After this make check fails on the OpenRISC port. > > tst-linkall-static fails with: > > make[3]: *** No rule to make target 'build/glibcs/or1k-linux-gnu-soft/glibc/resolv/libanl.a', needed by 'build/glibcs/or1k-linux-gnu-soft/glibc/elf/tst-linkall-static'. Stop. > > We may want to keep libanl.a around as an empty binary for build compatibility > reasons like we did with libutil. There may be builds wheck hard code -lanl, > maybe not as much as -lutil though. > > Never the less, this patch seems to fix it: I think should just remove the libanl.a reference, as per 699361795f6af8. I will fix it. > > diff --git a/elf/Makefile b/elf/Makefile > index 06bfa1642f..71bc2987df 100644 > --- a/elf/Makefile > +++ b/elf/Makefile > @@ -1683,7 +1683,6 @@ $(objpfx)tst-linkall-static: \ > $(common-objpfx)resolv/libresolv.a \ > $(common-objpfx)login/libutil.a \ > $(common-objpfx)rt/librt.a \ > - $(common-objpfx)resolv/libanl.a \ > $(static-thread-library) > > ifeq ($(build-crypt),yes) > @@ -1704,6 +1703,12 @@ $(objpfx)tst-linkall-static: \ > endif > endif > > +# Libanl is only available up to GLIBC_2.34 > +ifeq ($(have-GLIBC_2.34)$(have-thread-library),yesyes) > +$(objpfx)tst-linkall-static: \ > + $(common-objpfx)resolv/libanl.a > +endif > + > # The application depends on the DSO, and the DSO loads the plugin. > # The plugin also depends on the DSO. This creates the circular > # dependency via dlopen that we're testing to make sure works. > ^ permalink raw reply [flat|nested] 7+ messages in thread
* Re: [PATCH] resolv: Do not build libanl.so for ABIs starting at 2.35 2022-01-03 13:58 ` Adhemerval Zanella @ 2022-02-11 11:34 ` Stafford Horne 2022-02-11 11:42 ` Florian Weimer 0 siblings, 1 reply; 7+ messages in thread From: Stafford Horne @ 2022-02-11 11:34 UTC (permalink / raw) To: Adhemerval Zanella; +Cc: Florian Weimer, libc-alpha On Mon, Jan 03, 2022 at 10:58:53AM -0300, Adhemerval Zanella wrote: > > > On 03/01/2022 01:53, Stafford Horne wrote: > > On Thu, Dec 30, 2021 at 11:06:31AM -0300, Adhemerval Zanella via Libc-alpha wrote: > >> > >> > >> On 29/12/2021 15:30, Florian Weimer wrote: > >>> * Adhemerval Zanella: > >>> > >>>> --- > >>>> resolv/Makefile | 11 ++++++++++- > >>>> 1 file changed, 10 insertions(+), 1 deletion(-) > >>>> > >>>> diff --git a/resolv/Makefile b/resolv/Makefile > >>>> index 59e599535c..5606eab1fd 100644 > >>>> --- a/resolv/Makefile > >>>> +++ b/resolv/Makefile > >>>> @@ -78,8 +78,12 @@ generate := mtrace-tst-leaks.out tst-leaks.mtrace tst-leaks2.mtrace > >>>> > >>>> extra-libs := libresolv libnss_dns > >>>> ifeq ($(have-thread-library),yes) > >>>> -extra-libs += libanl > >>>> routines += gai_sigqueue > >>>> +endif > >>>> + > >>>> +ifeq ($(have-GLIBC_2.34)$(have-thread-library),yesyes) > >>>> +# Empty compatibility library for old binaries. > >>>> +extra-libs += libanl > >>>> > >>>> tests += \ > >>>> tst-bug18665 \ > >>>> @@ -176,6 +180,11 @@ $(libanl-routines-var) += \ > >>>> libanl-routines += libanl-compat > >>>> libanl-shared-only-routines += libanl-compat > >>>> > >>>> +# Pretend that libanl.so is a linker script, so that the symbolic link > >>>> +# is not installed. > >>>> +install-lib-ldscripts = libanl.so > >>>> +$(inst_libdir)/libanl.so: > >>>> + > >>> > >>> Looks okay. But the second hunk should be in a separate commit; I think > >>> we should backport it. > >> > >> Done. > > > > After this make check fails on the OpenRISC port. > > > > tst-linkall-static fails with: > > > > make[3]: *** No rule to make target 'build/glibcs/or1k-linux-gnu-soft/glibc/resolv/libanl.a', needed by 'build/glibcs/or1k-linux-gnu-soft/glibc/elf/tst-linkall-static'. Stop. > > > > We may want to keep libanl.a around as an empty binary for build compatibility > > reasons like we did with libutil. There may be builds wheck hard code -lanl, > > maybe not as much as -lutil though. > > > > Never the less, this patch seems to fix it: > > I think should just remove the libanl.a reference, as per 699361795f6af8. > I will fix it. One more thing, sorry I just picked this up now. After the fixes we are now getting: make: Entering directory '/home/shorne/work/gnu-toolchain/glibc/resolv' make: *** No rule to make target '/home/shorne/work/gnu-toolchain/build-glibc/resolv/tst-resolv-res_ninit.out', needed by '/home/shorne/work/gnu-toolchain/build-glibc/resolv/mtrace-tst-resolv-res_ninit.out'. Stop. make: Leaving directory '/home/shorne/work/gnu-toolchain/glibc/resolv' The "if" condition change blocked many of the tst-resolv* tests from being built or run for 2.35 only ports. Is that expected? There are now a few things like above that are missing their dependencies. Do we need: diff --git a/resolv/Makefile b/resolv/Makefile index c465479e8b..11e3b203a3 100644 --- a/resolv/Makefile +++ b/resolv/Makefile @@ -84,7 +84,9 @@ endif ifeq ($(have-GLIBC_2.34)$(have-thread-library),yesyes) # Empty compatibility library for old binaries. extra-libs += libanl +endif +ifeq ($(have-thread-library),yes) tests += \ tst-bug18665 \ tst-bug18665-tcp \ ^ permalink raw reply [flat|nested] 7+ messages in thread
* Re: [PATCH] resolv: Do not build libanl.so for ABIs starting at 2.35 2022-02-11 11:34 ` Stafford Horne @ 2022-02-11 11:42 ` Florian Weimer 0 siblings, 0 replies; 7+ messages in thread From: Florian Weimer @ 2022-02-11 11:42 UTC (permalink / raw) To: Stafford Horne; +Cc: Adhemerval Zanella, libc-alpha * Stafford Horne: > One more thing, sorry I just picked this up now. After the fixes we are now > getting: > > make: Entering directory '/home/shorne/work/gnu-toolchain/glibc/resolv' > make: *** No rule to make target '/home/shorne/work/gnu-toolchain/build-glibc/resolv/tst-resolv-res_ninit.out', needed by '/home/shorne/work/gnu-toolchain/build-glibc/resolv/mtrace-tst-resolv-res_ninit.out'. Stop. > make: Leaving directory '/home/shorne/work/gnu-toolchain/glibc/resolv' > > The "if" condition change blocked many of the tst-resolv* tests from being > built or run for 2.35 only ports. Is that expected? There are now a few > things like above that are missing their dependencies. Do we need: > > diff --git a/resolv/Makefile b/resolv/Makefile > index c465479e8b..11e3b203a3 100644 > --- a/resolv/Makefile > +++ b/resolv/Makefile > @@ -84,7 +84,9 @@ endif > ifeq ($(have-GLIBC_2.34)$(have-thread-library),yesyes) > # Empty compatibility library for old binaries. > extra-libs += libanl > +endif > > +ifeq ($(have-thread-library),yes) > tests += \ > tst-bug18665 \ > tst-bug18665-tcp \ Sorry about that, I should have reviewed this more closely. Can you move the libanl settings closer together? $(libanl-routines-var) should be outside the $(have-GLIBC_2.34) condition, libanl-routines += libanl-compat should be inside it. Thanks, Florian ^ permalink raw reply [flat|nested] 7+ messages in thread
end of thread, other threads:[~2022-02-11 11:43 UTC | newest] Thread overview: 7+ messages (download: mbox.gz / follow: Atom feed) -- links below jump to the message on this page -- 2021-12-29 16:39 [PATCH] resolv: Do not build libanl.so for ABIs starting at 2.35 Adhemerval Zanella 2021-12-29 18:30 ` Florian Weimer 2021-12-30 14:06 ` Adhemerval Zanella 2022-01-03 4:53 ` Stafford Horne 2022-01-03 13:58 ` Adhemerval Zanella 2022-02-11 11:34 ` Stafford Horne 2022-02-11 11:42 ` Florian Weimer
This is a public inbox, see mirroring instructions for how to clone and mirror all data and code used for this inbox; as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).