Mailing List Archive

On Tue, Apr 16, 2024 at 01:55:18PM -0400, Dennis Clarke via perl5-porters wrote:
>
> I see a strange error when I attempt to configure perl on a NetBSD 9.3
> server :
>
> .
> .
> .
> If this doesn't make any sense to you, just accept the default 'n'.
> Build a threading Perl? [n]
> Your platform has some specific hints regarding threaded builds, using
> them...
>
> Perl can be built so that multiple Perl interpreters can coexist
> within the same Perl executable.
>
> If this doesn't make any sense to you, just accept the default 'n'.
> Build Perl for multiplicity? [n]
>
> Use which C compiler? [cc] gcc
> gcc: error: unrecognized command line option ‘--whole-archive’; did you mean
> ‘--whole-program’?
>
> Hmm... Looks kind of like a BSD system, but we'll see...
>
> .
> .
> .
>
> I have never seen that before.
>
> Perhaps there is some strange compiler test in the config script?
>
>
> For the record this is NetBSD 9.3 on sparcv9 :
>
> p$
> p$ uname -a
> NetBSD pollux.bw.genunix.com 9.3_STABLE NetBSD 9.3_STABLE (GENERIC) #0: Tue
> Mar 12 12:45:42 UTC 2024
> mkrepro@mkrepro.NetBSD.org:/usr/src/sys/arch/sparc64/compile/GENERIC sparc64
> p$

There's a test done in the compiler callback unit to test where the
compiler supports that flag. It looks like it needs an extra
redirection to prevent the noise.

The "--whole-archive" option was needed in very old NetBSD, though
no-one involved (including myself, who added the CBU) added a ticket
reference.

Tony

On 4/16/24 23:40, Tony Cook wrote:
> On Tue, Apr 16, 2024 at 01:55:18PM -0400, Dennis Clarke via perl5-porters wrote:
>>
>> I see a strange error when I attempt to configure perl on a NetBSD 9.3
>> server :
>>
>> .
>> .
>> .
>> If this doesn't make any sense to you, just accept the default 'n'.
>> Build a threading Perl? [n]
>> Your platform has some specific hints regarding threaded builds, using
>> them...
>>
>> Perl can be built so that multiple Perl interpreters can coexist
>> within the same Perl executable.
>>
>> If this doesn't make any sense to you, just accept the default 'n'.
>> Build Perl for multiplicity? [n]
>>
>> Use which C compiler? [cc] gcc
>> gcc: error: unrecognized command line option ‘--whole-archive’; did you mean
>> ‘--whole-program’?
>>
>> Hmm... Looks kind of like a BSD system, but we'll see...
>>
>> .
>> .
>> .
>>
>> I have never seen that before.
>>
>> Perhaps there is some strange compiler test in the config script?
>>
>>
>> For the record this is NetBSD 9.3 on sparcv9 :
>>
>> p$
>> p$ uname -a
>> NetBSD pollux.bw.genunix.com 9.3_STABLE NetBSD 9.3_STABLE (GENERIC) #0: Tue
>> Mar 12 12:45:42 UTC 2024
>> mkrepro@mkrepro.NetBSD.org:/usr/src/sys/arch/sparc64/compile/GENERIC sparc64
>> p$
>
> There's a test done in the compiler callback unit to test where the
> compiler supports that flag. It looks like it needs an extra
> redirection to prevent the noise.
>
> The "--whole-archive" option was needed in very old NetBSD, though
> no-one involved (including myself, who added the CBU) added a ticket
> reference.
>
> Tony

Looks trivial to fix and I am working on it. The compile is running
now and I will report back when I have a clean testsuite.



--
--
Dennis Clarke
RISC-V/SPARC/PPC/ARM/CISC
UNIX and Linux spoken

On 4/16/24 23:40, Tony Cook wrote:
> On Tue, Apr 16, 2024 at 01:55:18PM -0400, Dennis Clarke via perl5-porters wrote:
>>
>> I see a strange error when I attempt to configure perl on a NetBSD 9.3
>> server :
>>
...
> The "--whole-archive" option was needed in very old NetBSD, though
> no-one involved (including myself, who added the CBU) added a ticket
> reference.
>
> Tony

OKay I seem to have exocised the demons.

A few trivial edits later and we get a clean compile :

pollux#
pollux# pwd
/opt/bw/build/perl-5.38.2
pollux#
pollux#
pollux# uname -a
NetBSD pollux.bw.genunix.com 9.3_STABLE NetBSD 9.3_STABLE (GENERIC) #0:
Tue Mar 12 12:45:42 UTC 2024
mkrepro@mkrepro.NetBSD.org:/usr/src/sys/arch/sparc64/compile/GENERIC sparc64
pollux#
pollux# pwd
/opt/bw/build/perl-5.38.2
pollux#
pollux# diff -u ./hints/netbsd.sh.orig ./hints/netbsd.sh
--- ./hints/netbsd.sh.orig 2023-11-28 11:57:28.000000000 +0000
+++ ./hints/netbsd.sh 2024-04-17 15:10:10.598422488 +0000
@@ -41,8 +41,8 @@
# system gcc to build correctly, so check for it
echo 'int f(void) { return 0; }' >try.c
if ${cc:-cc} $cccdlflags -c try.c -otry.o 2>&1 &&
- ${cc:-cc} --whole-archive $lddlflags try.o -otry.so 2>&1 ; then
- lddlflags="--whole-archive $lddlflags"
+ ${cc:-cc} $lddlflags try.o -otry.so 2>&1 ; then
+ lddlflags=" $lddlflags"
fi
rm try.c try.o try.so 2>/dev/null
EOCBU
@@ -55,8 +55,7 @@
# built as C++ objects (__eh_alloc, __pure_virtual,
# etc.) will always be defined.
#
- ccdlflags="-Wl,-whole-archive -lgcc \
- -Wl,-no-whole-archive -Wl,-E $ccdlflags"
+ ccdlflags=" -lgcc -Wl,-E $ccdlflags"
;;
*)
ccdlflags="-Wl,-E $ccdlflags"
pollux#
pollux# diff -u ./hints/freemint.sh.orig ./hints/freemint.sh
--- ./hints/freemint.sh.orig 2023-11-28 11:57:28.000000000 +0000
+++ ./hints/freemint.sh 2024-04-17 15:10:27.922457397 +0000
@@ -5,7 +5,7 @@
archname="m68k-freemint"

cccdlflags=' '; # avoid -fPIC
-ccdlflags="-Wl,-whole-archive -lgcc -lpthread -Wl,-no-whole-archive"
+ccdlflags="-lgcc -lpthread"

# libs
libpth="$prefix/lib /usr/local/lib /usr/lib"
pollux#
pollux# diff -u ./Porting/bisect-runner.pl.orig ./Porting/bisect-runner.pl
--- ./Porting/bisect-runner.pl.orig 2024-04-17 14:56:05.271872880 +0000
+++ ./Porting/bisect-runner.pl 2023-11-28 11:57:27.000000000 +0000
@@ -3028,7 +3028,7 @@
d_dlerror=$define
ccdlflags="-Wl,-E -Wl,-R${PREFIX}/lib $ccdlflags"
cccdlflags="-DPIC -fPIC $cccdlflags"
- lddlflags="-shared $lddlflags"
+ lddlflags="--whole-archive -shared $lddlflags"
elif [ "`uname -m`" = "pmax" ]; then
# NetBSD 1.3 and 1.3.1 on pmax shipped an 'old' ld.so, which will not
work.
d_dlopen=$undef
pollux#

Let me get that tested and report back.



--
Dennis Clarke
RISC-V/SPARC/PPC/ARM/CISC
UNIX and Linux spoken

Just interesting, why did you touch "fremont.sh" and "bisect-runner.pl" (in
addition)?

See also: https://github.com/Perl/perl5/pull/22152/files

Thanks

On Wed, Apr 17, 2024 at 8:04?PM Dennis Clarke via perl5-porters <
perl5-porters@perl.org> wrote:

> On 4/16/24 23:40, Tony Cook wrote:
> > On Tue, Apr 16, 2024 at 01:55:18PM -0400, Dennis Clarke via
> perl5-porters wrote:
> >>
> >> I see a strange error when I attempt to configure perl on a NetBSD 9.3
> >> server :
> >>
> ...
> > The "--whole-archive" option was needed in very old NetBSD, though
> > no-one involved (including myself, who added the CBU) added a ticket
> > reference.
> >
> > Tony
>
> OKay I seem to have exocised the demons.
>
> A few trivial edits later and we get a clean compile :
>
> pollux#
> pollux# pwd
> /opt/bw/build/perl-5.38.2
> pollux#
> pollux#
> pollux# uname -a
> NetBSD pollux.bw.genunix.com 9.3_STABLE NetBSD 9.3_STABLE (GENERIC) #0:
> Tue Mar 12 12:45:42 UTC 2024
> mkrepro@mkrepro.NetBSD.org:/usr/src/sys/arch/sparc64/compile/GENERIC
> sparc64
> pollux#
> pollux# pwd
> /opt/bw/build/perl-5.38.2
> pollux#
> pollux# diff -u ./hints/netbsd.sh.orig ./hints/netbsd.sh
> --- ./hints/netbsd.sh.orig 2023-11-28 11:57:28.000000000 +0000
> +++ ./hints/netbsd.sh 2024-04-17 15:10:10.598422488 +0000
> @@ -41,8 +41,8 @@
> # system gcc to build correctly, so check for it
> echo 'int f(void) { return 0; }' >try.c
> if ${cc:-cc} $cccdlflags -c try.c -otry.o 2>&1 &&
> - ${cc:-cc} --whole-archive $lddlflags try.o -otry.so 2>&1 ; then
> - lddlflags="--whole-archive $lddlflags"
> + ${cc:-cc} $lddlflags try.o -otry.so 2>&1 ; then
> + lddlflags=" $lddlflags"
> fi
> rm try.c try.o try.so 2>/dev/null
> EOCBU
> @@ -55,8 +55,7 @@
> # built as C++ objects (__eh_alloc,
> __pure_virtual,
> # etc.) will always be defined.
> #
> - ccdlflags="-Wl,-whole-archive -lgcc \
> - -Wl,-no-whole-archive -Wl,-E $ccdlflags"
> + ccdlflags=" -lgcc -Wl,-E $ccdlflags"
> ;;
> *)
> ccdlflags="-Wl,-E $ccdlflags"
> pollux#
> pollux# diff -u ./hints/freemint.sh.orig ./hints/freemint.sh
> --- ./hints/freemint.sh.orig 2023-11-28 11:57:28.000000000 +0000
> +++ ./hints/freemint.sh 2024-04-17 15:10:27.922457397 +0000
> @@ -5,7 +5,7 @@
> archname="m68k-freemint"
>
> cccdlflags=' '; # avoid -fPIC
> -ccdlflags="-Wl,-whole-archive -lgcc -lpthread -Wl,-no-whole-archive"
> +ccdlflags="-lgcc -lpthread"
>
> # libs
> libpth="$prefix/lib /usr/local/lib /usr/lib"
> pollux#
> pollux# diff -u ./Porting/bisect-runner.pl.orig ./Porting/bisect-runner.pl
> --- ./Porting/bisect-runner.pl.orig 2024-04-17 14:56:05.271872880
> +0000
> +++ ./Porting/bisect-runner.pl 2023-11-28 11:57:27.000000000 +0000
> @@ -3028,7 +3028,7 @@
> d_dlerror=$define
> ccdlflags="-Wl,-E -Wl,-R${PREFIX}/lib $ccdlflags"
> cccdlflags="-DPIC -fPIC $cccdlflags"
> - lddlflags="-shared $lddlflags"
> + lddlflags="--whole-archive -shared $lddlflags"
> elif [ "`uname -m`" = "pmax" ]; then
> # NetBSD 1.3 and 1.3.1 on pmax shipped an 'old' ld.so, which will not
> work.
> d_dlopen=$undef
> pollux#
>
> Let me get that tested and report back.
>
>
>
> --
> Dennis Clarke
> RISC-V/SPARC/PPC/ARM/CISC
> UNIX and Linux spoken
>
>

On 4/17/24 16:42, Elvin Aslanov wrote:
> Just interesting, why did you touch "fremont.sh" and "bisect-runner.pl" (in
> addition)?
>
> See also: https://github.com/Perl/perl5/pull/22152/files
I was trying to get perl to compile. That is why. I really do not care
where that "whole-archive" exists. I will nuke it from orbit to get a
clean compile. Which I did.

The testing process blows up real fast ... sadly :

/opt/bw/build/perl-5.38.2_netbsd_9.3_sparcv9.003/dist/threads/../../miniperl
"-I../../lib" -MExtUtils::Command::MM -e 'cp_nonempty' -- threads.bs
../../lib/auto/threads/threads.bs 644
./miniperl -Ilib make_ext.pl lib/auto/threads/shared/shared.so
MAKE="make" LIBPERL_A=libperl.a LINKTYPE=dynamic
/opt/bw/build/perl-5.38.2_netbsd_9.3_sparcv9.003/dist/threads-shared/../../miniperl
"-I../../lib" -MExtUtils::Command::MM -e 'cp_nonempty' -- shared.bs
../../lib/auto/threads/shared/shared.bs 644
cd t && (rm -f perl; /bin/ln -s ../perl perl)
./runtests choose
*** Error code 1

Stop.
make: stopped in /opt/bw/build/perl-5.38.2_netbsd_9.3_sparcv9.003

No clue what is going on there.





--
Dennis Clarke
RISC-V/SPARC/PPC/ARM/CISC
UNIX and Linux spoken