| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
|
|
| |
[amended to fix a typo, and add an exception to the podcheck database]
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
|
|
|
|
|
| |
The corresponding types are long long rather than __int64 so this makes
more sense and matches proposed diffs from the land of Strawberry Perl
for supporting 32-bit builds with 64-bit ints on Windows (hopefully
coming up soon...).
|
| |
|
|
|
| |
ExtUtils::MakeMaker and ExtUtils::CBuilder both default to this anyway,
but it's good to be explicit, and reduces diffs with the .gc64 file.
|
| | |
|
| |
|
|
| |
This entry is only for the user-visible changes.
|
| |
|
|
|
|
|
| |
As reported by Anša Vernerová:
| Calling the third argument $remainer is simply wrong - the remainder is
| the fourth argument, which is not assigned to anything.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
With the re_eval jumbo fix, the behaviour of overloaded objects in
runtime patterns, such /^$overloaded/ has changed, such that the stringify
overload ("") no longer avoids the need for 'use re "eval"': for example,
use overload "" => sub { qr/(??{1})/ }
my $o = bless [];
"1" =~ /^$o/;
works in 5.16.0, but dies with "Eval-group not allowed" in blead.
Change this back to the former behaviour, such that if qr and concat
ops aren't overloaded, then use "" overloading, and if the return from
that is a qr object, extract any code blocks from it.
This is achieved by:
* moving the concat/stringify code ahead of the regex block extraction
code,
* making the overloaded stringify call be explicit (rather than
being invoked implicitly by sv_catsv()),
* looping to re-apply overloading to any object returned by "".
* applying those last two steps in the case of a single arg too
This is a partial fix for
[perl #113682] Bleadperl v5.17.0-424-gd24ca0c breaks
ABIGAIL/Regexp-Common-2011121001.tar.gz
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
| |
Add examples showing how to use cflags.SH to tweak the compiler flags used
for individual object files.
Previously cflags.SH contained a somewhat stale pre-canned list of file
basenames including removed files such as usersub (deleted before 5.000
shipped), and a partial list of 5.000 XS extensions. Whilst it's possible
to generate the correct list in cflags by parsing MANIFEST (and adding a few
fixups), it's still not actually *useful*, as cflags gets overwritten as
soon as config.sh changes. Hence the most end-user useful solution with
minimal maintenance is to eliminate the list entirely, and document how the
user should add to it as necessary.
|
| |
|
|
|
|
| |
The previous code assumed that incrementing a signed integer value wraps.
We're lucky that it has (so far), as it's undefined behaviour in C.
So refactor to code which doesn't assume anything.
|
| |
|
|
|
|
| |
The code was treating any negative value as "ok", rather than only those
within 1e-9 of the expected value. This bug has been present since
tryeq_sloppy() was added by commit 800e6488b45c59f1 in Oct 2001.
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
|
|
| |
The CPAN release has an old version.
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
|
|
|
|
| |
moving files around doesn't change the mtime, and tarballs store the
mtime to use when untarring, so it's possible to end up with an updated
file with an mtime before the corresponding file in lib/
|
| |
|
|
|
|
| |
Filter::Util::Call comes from Filter-x.yz.tar.gz for instance, which
untars to Filter-x.yz, not Filter-Util-Call-x.yz as it was previously
assuming
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
|
|
|
| |
Storable.xs requires a definition of dVAR, which on 5.8.9 is only supplied
by ppport.h.
|
| |\ |
|
| | |
| |
| |
| |
| | |
As find_git_or_skip() no longer changes directory, no need to make paths in
@INC absolute in the various tests that use it.
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Return the directory of the source tree, so that tests can change directory
to it if necessary. Modify t/porting/cmp_version.t to change directory.
This gets t/porting/utils.t passing again under -Dmksymlinks, without breaking
t/porting/cmp_version.t. The former needs to run in the build tree as it
inspects various build products. The latter needs to run in the (real)
checkout tree, else it thinks that every single file has changed (into a
symlink).
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Previously these were overlooked - the generated .pm file corresponding to
an .xs file would be found on disk, but as that file was not in git, the
original $VERSION could not be determined and it would be skipped. Now the
correct master file is checked, and failed if it is not updated.
Without this change, t/porting/cmp_version.t currently fails for a build
using -Dmksymlinks, due to ext/DynaLoader/dl_vms.xs having recent changes.
The test changes directory to the original git checkout, and if that
checkout is clean, there is *no* ext/DynaLoader.pm file there, which meant
that the previous version of pm_file_from_xs() would die because it could
not figure out which Perl file relates to dl_vms.xs
|
| | | |
|
| | |
| |
| |
| |
| |
| |
| | |
For any changed module, if the old version is undefined or unparseable, treat
it as a SKIP, not an OK. If the old version *was* parseable, but the current
version is not, treat that as a FAIL - moving from a parseable to an
unparseable version is a regression.
|
