bug#62210: Job build-image-gnustep does not run anymore on EMBA

bug#62210: Job build-image-gnustep does not run anymore on EMBA

[Michael Albinus]

Hi,

Recently, we have upgraded EMBA jobs to run using Debian 11 (bullseye)
instead of Debian 9 (stretch). Since then, the gitlab job
build-image-gnustep fails. See for example
<https://emba.gnu.org/emacs/emacs/-/jobs/64723/raw>.

Best regards, Michael.

bug#62210: Job build-image-gnustep does not run anymore on EMBA

[Michael Albinus]

Michael Albinus <michael.albinus@gmx.de> writes:

Hi,

> Recently, we have upgraded EMBA jobs to run using Debian 11 (bullseye)
> instead of Debian 9 (stretch). Since then, the gitlab job
> build-image-gnustep fails. See for example
> <https://emba.gnu.org/emacs/emacs/-/jobs/64723/raw>.

Nobody has answered, so I comment out the tests temporarily.

Best regards, Michael.

bug#62210: Job build-image-gnustep does not run anymore on EMBA

[Michael Albinus]

Po Lu <luangruo@yahoo.com> writes:

Hi,

>> Since we have upgraded EMBA to Debian bullseye, the NS port tests
>> fail. See bug#62210.
>>
>> Alan Third, who has worked on this in the past, has stopped his
>> activities. Is there somebody else who cares now?
>
> The test times out because GCC keeps emitting massive amounts of
> warnings about non portable preprocessor constructs in GNUstep headers.
>
> Does this fix the problem:
>
> diff --git a/configure.ac b/configure.ac
> index c5300beb2ba..511199454e0 100644
> --- a/configure.ac
> +++ b/configure.ac
> @@ -2062,8 +2062,8 @@ AC_DEFUN
>         GNUSTEP_LOCAL_HEADERS="-I${GNUSTEP_LOCAL_HEADERS}"
>       test "x${GNUSTEP_LOCAL_LIBRARIES}" != "x" && \
>         GNUSTEP_LOCAL_LIBRARIES="-L${GNUSTEP_LOCAL_LIBRARIES}"
> -     CPPFLAGS="$CPPFLAGS -I${GNUSTEP_SYSTEM_HEADERS} ${GNUSTEP_LOCAL_HEADERS}"
> -     CFLAGS="$CFLAGS -I${GNUSTEP_SYSTEM_HEADERS} ${GNUSTEP_LOCAL_HEADERS}"
> +     CPPFLAGS="$CPPFLAGS -isystem ${GNUSTEP_SYSTEM_HEADERS} ${GNUSTEP_LOCAL_HEADERS}"
> +     CFLAGS="$CFLAGS -isystem ${GNUSTEP_SYSTEM_HEADERS} ${GNUSTEP_LOCAL_HEADERS}"
>       LDFLAGS="$LDFLAGS -L${GNUSTEP_SYSTEM_LIBRARIES} ${GNUSTEP_LOCAL_LIBRARIES}"
>       LIBS_GNUSTEP="-lgnustep-gui -lgnustep-base -lobjc -lpthread"
>       dnl GNUstep defines BASE_NATIVE_OBJC_EXCEPTIONS to 0 or 1.

Will try. Lets continue discussion on bug#62210.

Best regards, Michael.

bug#62210: Job build-image-gnustep does not run anymore on EMBA

[Michael Albinus]

Michael Albinus <michael.albinus@gmx.de> writes:

Hi,

>> The test times out because GCC keeps emitting massive amounts of
>> warnings about non portable preprocessor constructs in GNUstep headers.
>>
>> Does this fix the problem:
>>
>> diff --git a/configure.ac b/configure.ac
>> index c5300beb2ba..511199454e0 100644
>> --- a/configure.ac
>> +++ b/configure.ac
>> @@ -2062,8 +2062,8 @@ AC_DEFUN
>>         GNUSTEP_LOCAL_HEADERS="-I${GNUSTEP_LOCAL_HEADERS}"
>>       test "x${GNUSTEP_LOCAL_LIBRARIES}" != "x" && \
>>         GNUSTEP_LOCAL_LIBRARIES="-L${GNUSTEP_LOCAL_LIBRARIES}"
>> -     CPPFLAGS="$CPPFLAGS -I${GNUSTEP_SYSTEM_HEADERS} ${GNUSTEP_LOCAL_HEADERS}"
>> -     CFLAGS="$CFLAGS -I${GNUSTEP_SYSTEM_HEADERS} ${GNUSTEP_LOCAL_HEADERS}"
>> +     CPPFLAGS="$CPPFLAGS -isystem ${GNUSTEP_SYSTEM_HEADERS} ${GNUSTEP_LOCAL_HEADERS}"
>> +     CFLAGS="$CFLAGS -isystem ${GNUSTEP_SYSTEM_HEADERS} ${GNUSTEP_LOCAL_HEADERS}"
>>       LDFLAGS="$LDFLAGS -L${GNUSTEP_SYSTEM_LIBRARIES} ${GNUSTEP_LOCAL_LIBRARIES}"
>>       LIBS_GNUSTEP="-lgnustep-gui -lgnustep-base -lobjc -lpthread"
>>       dnl GNUstep defines BASE_NATIVE_OBJC_EXCEPTIONS to 0 or 1.
>
> Will try. Lets continue discussion on bug#62210.

That didn't help. See <https://emba.gnu.org/emacs/emacs/-/jobs/65167>.

Best regards, Michael.

bug#62210: Job build-image-gnustep does not run anymore on EMBA

[Michael Albinus]

Po Lu <luangruo@yahoo.com> writes:

Hi,

>> That didn't help. See <https://emba.gnu.org/emacs/emacs/-/jobs/65167>.
>
> Would it be possible to do `make V=1' and show exactly how GCC is being
> called?

Will do.

Best regards, Michael.

bug#62210: Job build-image-gnustep does not run anymore on EMBA

[Michael Albinus]

Michael Albinus <michael.albinus@gmx.de> writes:

> Po Lu <luangruo@yahoo.com> writes:

Hi,

>>> That didn't help. See <https://emba.gnu.org/emacs/emacs/-/jobs/65167>.
>>
>> Would it be possible to do `make V=1' and show exactly how GCC is being
>> called?
>
> Will do.

<https://emba.gnu.org/emacs/emacs/-/jobs/65276/raw>

Best regards, Michael.

bug#62210: Status uf NS port?

[Robert Pluim]

>>>>> On Tue, 04 Apr 2023 13:05:18 +0200, Michael Albinus <michael.albinus@gmx.de> said:

    Michael> Alan Third <alan@idiocy.org> writes:
    Michael> Hi Alan,

    >> I never managed to work out how to build against gnustep with GCC.
    >> Clang worked absolutely fine.
    >> 
    >> FWIW, gnustep != macOS. Gnustep has very few users, and in all the
    >> years the gnustep port of Emacs was entirely broken we didn't receive
    >> a single bug report about it. Except maybe from Martin, but he was
    >> only using it as a test bed.

    Michael> Thanks for the feedback. I've commented out the GNUSTEP jobs on EMBA
    Michael> (again), until somebody provides running build instructions.

If the issue is the amount of warnings produced by the build, the
following reduces those to almost nothing on a '--with-ns' build on
Debian testing.

diff --git a/src/nsgui.h b/src/nsgui.h
index 469aef13070..3b977e531a5 100644
--- a/src/nsgui.h
+++ b/src/nsgui.h
@@ -31,8 +31,20 @@ #define Cursor FooFoo
 
 #undef verify
 
+#ifdef NS_IMPL_GNUSTEP
+#pragma GCC diagnostic push
+#pragma GCC diagnostic ignored "-Wexpansion-to-defined"
+#pragma GCC diagnostic ignored "-Wold-style-definition"
+#pragma GCC diagnostic ignored "-Wstrict-prototypes"
+#pragma GCC diagnostic ignored "-Wredundant-decls"
+#endif
+
 #import <AppKit/AppKit.h>
 
+#ifdef NS_IMPL_GNUSTEP
+#pragma GCC diagnostic pop
+#endif
+
 #ifdef NS_IMPL_COCOA
 #undef Cursor
 #endif /* NS_IMPL_COCOA */

bug#62210: Status uf NS port?

[Michael Albinus]

Robert Pluim <rpluim@gmail.com> writes:

Hi Robert,

> If the issue is the amount of warnings produced by the build, the
> following reduces those to almost nothing on a '--with-ns' build on
> Debian testing.
>
> diff --git a/src/nsgui.h b/src/nsgui.h
> index 469aef13070..3b977e531a5 100644
> --- a/src/nsgui.h
> +++ b/src/nsgui.h
> @@ -31,8 +31,20 @@ #define Cursor FooFoo
>
>  #undef verify
>
> +#ifdef NS_IMPL_GNUSTEP
> +#pragma GCC diagnostic push
> +#pragma GCC diagnostic ignored "-Wexpansion-to-defined"
> +#pragma GCC diagnostic ignored "-Wold-style-definition"
> +#pragma GCC diagnostic ignored "-Wstrict-prototypes"
> +#pragma GCC diagnostic ignored "-Wredundant-decls"
> +#endif
> +
>  #import <AppKit/AppKit.h>
>
> +#ifdef NS_IMPL_GNUSTEP
> +#pragma GCC diagnostic pop
> +#endif
> +
>  #ifdef NS_IMPL_COCOA
>  #undef Cursor
>  #endif /* NS_IMPL_COCOA */

Thanks, this reduces the warnings drastically. What's left are

--8<---------------cut here---------------start------------->8---
 CC       image.o
image.c:143: warning: macro "PIX_MASK_DRAW" is not used [-Wunused-macros]
  143 | #define PIX_MASK_DRAW 1
      |
  CC       nsterm.o
cc1obj: warning: /root/GNUstep/Library/Headers: No such file or directory [-Wmissing-include-dirs]
cc1obj: warning: /usr/local/include/GNUstep: No such file or directory [-Wmissing-include-dirs]
nsterm.m: In function 'ns_set_undecorated':
nsterm.m:1791:7: warning: 'NSWindow' may not respond to '-setIsVisible:'
 1791 |       [newWindow setIsVisible:[oldWindow isVisible]];
      |       ^
nsterm.m:1791:7: warning: (messages without a matching method signature will be assumed to return 'id' and accept '...' as arguments)
nsterm.m: In function '-[EmacsView showFontPanel]':
nsterm.m:6688:3: warning: 'NSFontPanel' may not respond to '-setIsVisible:'
 6688 |   [[fm fontPanel: YES] setIsVisible: NO];
      |   ^
  CC       nsfns.o
cc1obj: warning: /root/GNUstep/Library/Headers: No such file or directory [-Wmissing-include-dirs]
cc1obj: warning: /usr/local/include/GNUstep: No such file or directory [-Wmissing-include-dirs]
  CC       nsmenu.o
cc1obj: warning: /root/GNUstep/Library/Headers: No such file or directory [-Wmissing-include-dirs]
cc1obj: warning: /usr/local/include/GNUstep: No such file or directory [-Wmissing-include-dirs]
  CC       nsselect.o
cc1obj: warning: /root/GNUstep/Library/Headers: No such file or directory [-Wmissing-include-dirs]
cc1obj: warning: /usr/local/include/GNUstep: No such file or directory [-Wmissing-include-dirs]
  CC       nsimage.o
cc1obj: warning: /root/GNUstep/Library/Headers: No such file or directory [-Wmissing-include-dirs]
cc1obj: warning: /usr/local/include/GNUstep: No such file or directory [-Wmissing-include-dirs]
  CC       nsfont.o
cc1obj: warning: /root/GNUstep/Library/Headers: No such file or directory [-Wmissing-include-dirs]
cc1obj: warning: /usr/local/include/GNUstep: No such file or directory [-Wmissing-include-dirs]
--8<---------------cut here---------------end--------------->8---

I have no idea how serious these warnings are. But the build runs
through now.

I don't know whether we shall commit the patch, it suppresses warnings
by default, which could be dangerous either. I'll let this to
somebody else to decide. Eli?

The good message is, that I could identify now the real problem:

--8<---------------cut here---------------start------------->8---
/usr/bin/ld: cannot find -lz
collect2: error: ld returned 1 exit status
make[3]: *** [Makefile:705: temacs] Error 1
make[2]: *** [Makefile:544: src] Error 2
make[1]: *** [Makefile:1260: actual-bootstrap] Error 2
--8<---------------cut here---------------end--------------->8---

zlib is missing. I've added zlib1g-dev to the list of installation
packages for Debian, and now the GNUSTEP jobs shall run on EMBA.

Best regards, Michael.

bug#62210: Status uf NS port?

[Robert Pluim]

>>>>> On Tue, 04 Apr 2023 20:03:34 +0200, Michael Albinus <michael.albinus@gmx.de> said:

    Michael> I don't know whether we shall commit the patch, it suppresses warnings
    Michael> by default, which could be dangerous either. I'll let this to
    Michael> somebody else to decide. Eli?

Only for the appkit headers, not globally.

    Michael> The good message is, that I could identify now the real problem:

    Michael> /usr/bin/ld: cannot find -lz
    Michael> collect2: error: ld returned 1 exit status
    Michael> make[3]: *** [Makefile:705: temacs] Error 1
    Michael> make[2]: *** [Makefile:544: src] Error 2
    Michael> make[1]: *** [Makefile:1260: actual-bootstrap] Error 2

    Michael> zlib is missing. I've added zlib1g-dev to the list of installation
    Michael> packages for Debian, and now the GNUSTEP jobs shall run on EMBA.

Letʼs hope thatʼs enough.

Robert
-- 

bug#62210: Job build-image-gnustep does not run anymore on EMBA

[Stefan Kangas]

Robert Pluim <rpluim@gmail.com> writes:

>>>>>> On Tue, 04 Apr 2023 20:03:34 +0200, Michael Albinus <michael.albinus@gmx.de> said:
>
>     Michael> I don't know whether we shall commit the patch, it suppresses warnings
>     Michael> by default, which could be dangerous either. I'll let this to
>     Michael> somebody else to decide. Eli?
>
> Only for the appkit headers, not globally.
>
>     Michael> The good message is, that I could identify now the real problem:
>
>     Michael> /usr/bin/ld: cannot find -lz
>     Michael> collect2: error: ld returned 1 exit status
>     Michael> make[3]: *** [Makefile:705: temacs] Error 1
>     Michael> make[2]: *** [Makefile:544: src] Error 2
>     Michael> make[1]: *** [Makefile:1260: actual-bootstrap] Error 2
>
>     Michael> zlib is missing. I've added zlib1g-dev to the list of installation
>     Michael> packages for Debian, and now the GNUSTEP jobs shall run on EMBA.
>
> Letʼs hope thatʼs enough.

Was that all, or is there more to do here?  Thanks.

bug#62210: Job build-image-gnustep does not run anymore on EMBA

[Michael Albinus]

Stefan Kangas <stefankangas@gmail.com> writes:

Hi stefan,

> Robert Pluim <rpluim@gmail.com> writes:
>
>>>>>>> On Tue, 04 Apr 2023 20:03:34 +0200, Michael Albinus <michael.albinus@gmx.de> said:
>>
>>     Michael> I don't know whether we shall commit the patch, it suppresses warnings
>>     Michael> by default, which could be dangerous either. I'll let this to
>>     Michael> somebody else to decide. Eli?
>>
>> Only for the appkit headers, not globally.
>>
>>     Michael> The good message is, that I could identify now the real problem:
>>
>>     Michael> /usr/bin/ld: cannot find -lz
>>     Michael> collect2: error: ld returned 1 exit status
>>     Michael> make[3]: *** [Makefile:705: temacs] Error 1
>>     Michael> make[2]: *** [Makefile:544: src] Error 2
>>     Michael> make[1]: *** [Makefile:1260: actual-bootstrap] Error 2
>>
>>     Michael> zlib is missing. I've added zlib1g-dev to the list of installation
>>     Michael> packages for Debian, and now the GNUSTEP jobs shall run on EMBA.
>>
>> Letʼs hope thatʼs enough.
>
> Was that all, or is there more to do here?  Thanks.

For emba, that's OK. I've left the bug open in case somebody else wants
to work on gnustep. Looks like nobody is interested, so I'm closing the
bug.

Best regards, Michael.