Mojca Miklavec mojca at
Wed Jul 23 23:48:27 PDT 2014

Dear Frank,

first of all, thanks a lot for doing the mass update of the number of
Perl modules.

On Thu, Jul 24, 2014 at 12:26 AM, Frank Schima wrote:
> On Jul 23, 2014, at 9:20 AM, Mojca Miklavec wrote:
>> One slightly easier step-by-step approach could be the following
>> 1.) Replace
>>    perl5.branches      5.14 5.16 5.18
>> in all perl modules with
>>    perl5.branches_blacklist 5.8 5.10 5.12
>> or to remove branche_blacklist altogether if the module builds with
>> all perl versions. When that blacklist isn't present (or empty) simply
>> make the module work with all known perl versions (defined in the
>> PortGroup).
>> That would get us all the modules updated to 5.18 and 5.20 "almost for
>> free" (that change should probably be scripted to cover all thousand
>> ports).
> FWIW, I prefer this approach. This allows us to move forward faster.

Do you know how to implement this, possibly together with replaced_by
for 5.8 and 5.10? (If we would implement this, I would suggest to try
to get rid of 5.8 and 5.10 already in the first round.)

I'm annoyed by a bug in 5.10
which causes a whole cascade of perl modules to fail and that only
gives false alarms from failed runs on the buildbots when modifying

>> 3.) Make sure that no p5.12 is hardcoded anywhere (the ticket that has
>> recently been opened), maybe try to hardcode 5.20 instead of 5.16
>> where it works (or add variants).
> I was wondering if we could have a variable in base that is the default perl version.

That would also be nice. But we would need to make sure to revbump all
such ports anyway.

>> 5.) Once the above is achieved and working well for about a month or
>> more, we could in principle also remove 5.16 and 5.18. But that is not
>> absolutely necessary. In any case it would be a lot easier and safer
>> to do it stepwise than to simply jump to the 5.20 train and suffer
>> headaches from all the intermediate problems.
> As long as everything works in 5.20.

Yes, that's what I said.

> So far I’m hitting a few bumps with p5.20-digest-sha1 [1] and p5.18-pdl not compiling. I can’t even try p5.20-pdl until p5.20-digest-sha1 is fixed.

The digest-sha1 works for me, pdl takes a bit longer, but eventually
fails for me as well.

:info:build pdlcore.c:875:66: warning: format specifies type 'int' but
the argument has type 'PDL_Indx' (aka 'long') [-Wformat]
:info:build     fprintf(stderr,"pdl_kludge_copy: level=%d;
:info:build                                                      ~~
:info:build                                                      %ld
:info:build pdlcore.c:876:161: warning: format specifies type 'int'
but the argument has type 'long' [-Wformat]
:info:build     croak("Internal error - please submit a bug report at\n  pdl_kludge_copy: Assertion
failed; ndims-1-level (%d) < 0!.",ndims-1-level);

                   ~~         ^~~~~~~~~~~~~



Odd number of elements in hash assignment at
line 41.
Use of uninitialized value in join or string at
line 59.

The first one is
    my ( $sym, %hash ) = split chr(0), $txt;
and the second one is a consequence of the first one failing. I don't
speak Perl well enough to be able to diagnose.

I see that the buildbots failed to compile opengl and wx. OpenGL failed with:

DEBUG: Assembled command: 'cd
&& /opt/local/bin/perl5.18 Makefile.PL INSTALLDIRS=vendor
CC="/usr/bin/clang" LD="/usr/bin/clang"'
DEBUG: Executing command line:  cd
&& /opt/local/bin/perl5.18 Makefile.PL INSTALLDIRS=vendor
CC="/usr/bin/clang" LD="/usr/bin/clang"
_RegisterApplication(), FAILED TO establish the default connection to
the WindowServer, _CGSDefaultConnection() is NULL.
2014-07-23 15:04:56.241 glversion[9341:407] invalid CoreGraphics connection
2014-07-23 15:04:56.242 glversion[9341:407] invalid CoreGraphics connection
2014-07-23 15:04:56.242 glversion[9341:407] GLUT Fatal Error: pixel
format with necessary capabilities not found.
make: *** [glversion.txt] Error 1

but it works when I build it from source.


More information about the macports-dev mailing list