| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| |
|
|
| |
perl
|
| |
|
|
|
| |
mkppport --clean relies on miniperl not yet being deleted, which doesn't always
hold. Idea blatently stolen from Steve Hay's commit 85741d00cbd04edb.
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
|
|
| |
Make it warn that it is deprecated. To be removed once the smokers upgrade.
|
| |
|
|
|
|
|
|
|
|
|
|
| |
But not in the way it was done before with a recursive call to $(MAKE) -
instead create it directly in the current make invocation as an action for
test_prep, and create the link for perl.third directly as an action for
test_prep.third
Also, make it optional to pass the perl to run into runtests, by providing a
default of ./perl$(EXE_EXT)
This feels simpler and still Don't Repeat Yourself.
|
| |
|
|
|
| |
This avoids a recursive call to $(MAKE) in the top level directory, which feels
over-complicated when all the targets invoked have no dependencies.
|
| |
|
|
|
| |
'$(LDLIBPTH) $(RUN) ./miniperl$(EXE_EXT) -Ilib' becomes '$(MINIPERL)'
'miniperl$(EXE_EXT)' becomes '$(MINIPERL_EXE)'
|
| |
|
|
|
|
|
|
| |
lib.
In fact, I don't see why they need to chdir at all. Or still need most of the
self-extraction logic, now that we have MakeMaker. (Yes, the code is *that*
old.)
|
| |
|
|
| |
This makes generating the -I options for the command line clearer.
|
| |
|
|
|
|
|
|
|
| |
It was only parallel make order roulette that it was working on my machine.
VMS doesn't have this problem, as it's invoking mktables from the top level,
using its -C option to change directory, and VMS's $(MINIPERL) has an -I option
for [.ext.Cwd.lib]
Maybe all platforms should swap to using the -C approach.
|
| |
|
|
|
|
|
|
|
|
| |
It can't really be renamed from ext/Cwd to ext/PathTools, because Configure and
Makefile.SH need to know the name of the shared object produced, and they infer
this from the name of the directory.
This concludes the migration of modules from lib to ext. Exporter and version
remain in lib, but I don't think that there is any benefit in trying to move
either of them.
|
| |
|
|
|
| |
Nothing up to and including minitest uses lib, and not needing many special
case rules to extract from lib/lib_pm.PL simplifies all the Makefiles.
|
| |
|
|
|
|
|
| |
But not the tests yet.
We need to avoid running a sitecustomize script as it may have side effects
that cause the build to fail, or worse, complete but with buggy side effects.
|
| | |
|
| | |
|
| |
|
|
|
| |
Obviously, it's going to take a bit more work to piece the PathTools
distribution back together.
|
| | |
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
It seems that the code in the Makefile to run AutoSplit over lib has actually
been vestigial since some point *before* 5.000. Presumably, initially, Perl code
for extensions was in lib, and autosplit (the precursor to AutoSplit.pm) was run
on them there. Or they were copied from ext to lib first, before splitting.
Fifteen year-old backup tapes refuse to give us the detailed history of the
pre-5.000 period.
Whatever the cause, the code was never needed - backed up by the absense of any
analgous code in the Win32 Makefiles. So remove it. This happens to also remove
the only use of the shell script makedir, so remove it too.
|
| | |
|
| | |
|
| |
|
|
|
| |
(d1a2168615d3 isn't enough to ensure that the toolchain is in lib/ before we
start processing XS code.)
|
| |
|
|
|
|
|
|
| |
(It requires several modules that we'd like to move from lib/ to ext/
In turn, as it runs before the dynamic extensions are built, it means that
all modules *they* require at configure and build time will already be in lib,
which simplifies things considerably. We then only need to worry about a
subset of pure-perl modules in ext/ that require each other to build.)
|
| |
|
|
|
|
| |
This is no longer needed after reverting 4f87fa576f174a31e6d8fdb08cae76e644395db0.
This reverts commit 8d6e18f2d838bb62bb3431d103b145a05cf94e4a.
|
| |
|
|
| |
A Makefile hack courtesy of Nicholas.
|
| |
|
|
|
| |
Add hack to Makefile.SH to specify an explicit dependency of
Text::ParseWords on Scalar::Util.
|
| | |
|
| |
|
|
| |
Should really have been done as part of 344af494c35a9f0f50dab51474b2e7cd806f1b08
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
(The table is 256 bytes; the run-time initialisation code is larger than this!)
Adapt generate_uudmap.c to generate the initalisation block for PL_bitcount,
writing the code to bitcount.h, using the same approach as uudmap.h.
To preserve binary compatibility:
for MULTIPLICITY:
keep Ibitcount in the interpreter structure, but remove all the macros that
access it. PL_bitcount is a new symbol in the object file, which won't clash
with anything as that name wasn't used before.
otherwise:
keep PL_bitcount as a char *, but initialise it at compile time to a new
constant array PL_bitcount array. Remove the code that attempts to Safefree()
it at interpreter destruction time.
|
| |
|
|
| |
instead of having the calling Makefile redirect stdout.
|
| |
|
|
| |
(A perl build with a shared libperl needs this.)
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
|
|
| |
Add perldelta.pod to the generated_pods macro so that it is deleted still.
|
| | |
|
| | |
|
| | |
|
| |
|
|
|
|
| |
(Static and dynamic each build a target in lib/auto, which was unchanged.
nonxs_ext use a target inside the extension's directory, which was moved, but
Makefile.SH had not been updated to reflect this.)
|
| | |
|
| |
|
|
| |
(It was already in a different target that caused make clean to delete it.)
|
| |
|
|
|
|
| |
is a race condition which loses if configpm happens to test lib/Config.pm just
at the point where lib/Config_git.pl is freshly created, but has no content, and
hence returns false. So make it an explicit dependency of lib/Config.pm
|
