Fails to build with GCC 15

[sebastic]

As reported in Debian Bug #1097248:

The full build log can be found at:
http://qa-logs.debian.net/2025/02/16/amd64exp/libpdl-io-hdf5-perl_0.761-1_unstable_gccexp.log.gz
The last lines of the build log are at the end of this report.

To build with GCC 15, either set CC=gcc-15 CXX=g++-15 explicitly,
or install the gcc, g++, gfortran, ... packages from experimental.

apt-get -t=experimental install g++  

GCC 15 now defaults to the C23/C++23 standards, exposing many FTBFS.
Other common build failures are new warnings resulting in build failures
with -Werror turned on, or new/dropped symbols in Debian symbols files.
For other C/C++ related build failures see the porting guide at
http://gcc.gnu.org/gcc-15/porting_to.html

[...]
make[1]: Entering directory '/build/reproducible-path/libpdl-io-hdf5-perl-0.761'
"/usr/bin/perl" -MExtUtils::Command::MM -e 'cp_nonempty' -- HDF5.bs blib/arch/auto/PDL/IO/HDF5/HDF5.bs 644
PERL_DL_NONLAZY=1 "/usr/bin/perl" "-MExtUtils::Command::MM" "-MTest::Harness" "-e" "undef *Test::Harness::Switches; test_harness(1, 'blib/lib', 'blib/arch')" t/*.t
lib/PDL/PP.pm: loadable library and perl binaries are mismatched (got first handshake key 0xf200080, needed 0xeb80080)
t/attribPDL.t ...
Dubious, test returned 1 (wstat 256, 0x100)
No subtests run
lib/PDL/PP.pm: loadable library and perl binaries are mismatched (got first handshake key 0xf200080, needed 0xeb80080)
t/file.t ........
Dubious, test returned 1 (wstat 256, 0x100)
No subtests run
lib/PDL/PP.pm: loadable library and perl binaries are mismatched (got first handshake key 0xf200080, needed 0xeb80080)
t/group.t .......
Dubious, test returned 1 (wstat 256, 0x100)
No subtests run
lib/PDL/PP.pm: loadable library and perl binaries are mismatched (got first handshake key 0xf200080, needed 0xeb80080)
t/reference.t ...
Dubious, test returned 1 (wstat 256, 0x100)
No subtests run
lib/PDL/PP.pm: loadable library and perl binaries are mismatched (got first handshake key 0xf200080, needed 0xeb80080)
t/total.t .......
Dubious, test returned 1 (wstat 256, 0x100)
No subtests run
lib/PDL/PP.pm: loadable library and perl binaries are mismatched (got first handshake key 0xf200080, needed 0xeb80080)
t/unlink.t ......
Dubious, test returned 1 (wstat 256, 0x100)
No subtests run
lib/PDL/PP.pm: loadable library and perl binaries are mismatched (got first handshake key 0xf200080, needed 0xeb80080)
t/vlenString.t ..
Dubious, test returned 1 (wstat 256, 0x100)
No subtests run
lib/PDL/PP.pm: loadable library and perl binaries are mismatched (got first handshake key 0xf200080, needed 0xeb80080)
t/xData.t .......
Dubious, test returned 1 (wstat 256, 0x100)
No subtests run

Test Summary Report
-------------------
t/attribPDL.t (Wstat: 256 (exited 1) Tests: 0 Failed: 0)
  Non-zero exit status: 1
  Parse errors: No plan found in TAP output
t/file.t      (Wstat: 256 (exited 1) Tests: 0 Failed: 0)
  Non-zero exit status: 1
  Parse errors: No plan found in TAP output
t/group.t     (Wstat: 256 (exited 1) Tests: 0 Failed: 0)
  Non-zero exit status: 1
  Parse errors: No plan found in TAP output
t/reference.t (Wstat: 256 (exited 1) Tests: 0 Failed: 0)
  Non-zero exit status: 1
  Parse errors: No plan found in TAP output
t/total.t     (Wstat: 256 (exited 1) Tests: 0 Failed: 0)
  Non-zero exit status: 1
  Parse errors: No plan found in TAP output
t/unlink.t    (Wstat: 256 (exited 1) Tests: 0 Failed: 0)
  Non-zero exit status: 1
  Parse errors: No plan found in TAP output
t/vlenString.t (Wstat: 256 (exited 1) Tests: 0 Failed: 0)
  Non-zero exit status: 1
  Parse errors: No plan found in TAP output
t/xData.t     (Wstat: 256 (exited 1) Tests: 0 Failed: 0)
  Non-zero exit status: 1
  Parse errors: No plan found in TAP output
Files=8, Tests=0,  1 wallclock secs ( 0.06 usr  0.02 sys +  0.95 cusr  0.18 csys =  1.21 CPU)
Result: FAIL
Failed 8/8 test programs. 0/0 subtests failed.
make[1]: *** [Makefile:1065: test_dynamic] Error 1
make[1]: Leaving directory '/build/reproducible-path/libpdl-io-hdf5-perl-0.761'
dh_auto_test: error: make -j8 test TEST_VERBOSE=1 returned exit code 2
make: *** [debian/rules:15: binary] Error 25
dpkg-buildpackage: error: debian/rules binary subprocess returned exit status 2

[mohawk2]

Thanks for the report. The message says "lib/PDL/PP.pm: loadable library and perl binaries are mismatched (got first handshake key 0xf200080, needed 0xeb80080)" - what is PDL and/or PDL::IO::HDF5 supposed to do about this? It looks like the build environment is broken.

[sebastic]

The GCC change likely affects PDL::PP, because the build succeeds with GCC 14. After upgrading GCC to 15 the build fails with no changes in Perl or PDL.

[mohawk2]

I will copy-paste the same part of the log again: "lib/PDL/PP.pm: loadable library and perl binaries are mismatched (got first handshake key 0xf200080, needed 0xeb80080)".

Can you confirm the perl -V of the Perl that is being run there, and that it was compiled with the same GCC that PDL was? This build/test (of a PDL-depending module) relies (to state the obvious) on both Perl, and PDL, and those two have to match.

[sebastic]

Perl and PDL were not built with GCC 15 obviously:

# perl -V | grep gccversion
    gccversion='14.2.0'
# pdl -V | grep gccversion
    gccversion='14.2.0'

Is the conclusion then that perl, pdl, and libpdl-io-hdf5-perl all need to be built with the same GCC version?

[mohawk2]

Is the conclusion then that perl, pdl, and libpdl-io-hdf5-perl all need to be built with the same GCC version?

It's my surmise that's the case, yes. If so, that will mean GCC 14 -> 15 will have ABI break implications for transitioning.

[sebastic]

Niko Tyni in the Debian bugreport:

This is specific to libpdl-io-hdf5-perl and libpdl-netcdf-perl, not a
wider issue. It is caused by them adding -Dbool=int to the build flags.
I doubt it would work even if perl, pdl, and libpdl-io-hdf5-perl were
all built with GCC 15 (but I didn't bother to try.)

The Perl headers #include <stdbool.h> if available, which is the case
on Debian. This system header #defines bool only on C17 and lower, as
it has become a keyword of its own on C23. GCC 15 changes the default
C language version from C17 to C23.

The combined effect is that for C17 (gcc -std=gnu17, default on GCC 14)
the stdbool.h definition overrides the -Dbool=int command line definition,
rendering it harmless (if nonsensical). But for C23 (gcc -std=gnu23,
default on GCC 15) -Dbool=int actually changes the bool size, breaking
the XS module interface.

My suggestion is to remove the -Dbool=int fiddling from the respective
Makefile.PL files, at least on Debian. The PDL-NetCDF git history suggests
that it was added to fix a build problem on Red Hat Linux 5.2 with perl
5.00404, so presumably 1998 or so.

[mohawk2]

That makes sense. Are you content to make a patch just removing that bit of the Makefile.PL?

[sebastic]

Confirmed fixed with this patch:

Description: Don't set -Dbool=int to fix FTFBS with GCC 15.
Author: Bas Couwenberg <[email protected]>
Bug: https://github.com/PDLPorters/pdl-io-hdf5/issues/7
Bug-Debian: https://bugs.debian.org/1097248

--- a/Makefile.PL
+++ b/Makefile.PL
@@ -94,15 +94,6 @@ if( -e "$hdf5_include_path/H5config.h"){
        }
 }
 
-# The following code was originally in the PDL::netCDF Makefile.PL
-#  (Not sure if it is really needed here)
-# Check if compiled under gcc/Linux.  In which case, define bool for the compiler
-my $define_bool = '';
-if ($Config{'osname'} =~ /linux/) {
-  $define_bool = '-Dbool=int';
-  print "Defining bool=int (linux seems to need this)\n";
-}
-
 #If in win32, add the required defined for the HDF5 libs to work:
 my $define_win32HDF = '';
 if ($Config{'osname'} =~ /win32/i) {
@@ -176,7 +167,7 @@ my $preop = '$(PERL) -I$(INST_ARCHLIB) -
 
 WriteMakefile(
   'NAME'            => 'PDL::IO::HDF5',
-  'CCFLAGS'      => "$Config{ccflags} $define_bool $define_win32HDF -DH5_USE_16_API -g",
+  'CCFLAGS'      => "$Config{ccflags} $define_win32HDF -DH5_USE_16_API -g",
   'CONFIGURE_REQUIRES' => { PDL => '2.004' },
   'BUILD_REQUIRES'     => { PDL => '2.004' },
    # 'TEST_REQUIRES'      => { PDL => '2.004' },

[mohawk2]

Done similar here. Could you try it as a patch and if it works, close this?

[sebastic]

The changes from #8 are already included in the Debian package, those are the same as in edbfe8d.