summaryrefslogtreecommitdiff
path: root/write_buildcustomize.pl
Commit message (Collapse)AuthorAgeFilesLines
* Add Term::ReadLine to write_buildcustomize.pl.Craig A. Berry2011-12-021-0/+3
| | | | | | | | | | | | | | If the build fails while building extensions, it's nice to have the debugger available to help figure out what went wrong. You couldn't do that before because lib/perl5db.pl depends on Term::ReadLine, which wouldn't be available since it hadn't been built yet. This commit makes Term::ReadLine available via the same mechanism that makes other libraries available to miniperl during the build. An alternative would be to remove the debugger's dependency on Term::ReadLine, but that would be more work and more risk for a situation that hopefully doesn't come up that often.
* dual-life CarpZefram2011-09-041-1/+1
| | | | | | | | | Make Carp portable to older Perl versions: * check minimum Perl version (5.6) at load time * use || instead of // * attempt downgrading to avoid loading Unicode tables when that might fail * check whether utf8::is_utf8() exists before calling it * lower IPC::Open3 version requirement in Carp tests
* move Carp to ext/Carp, preparing for dual-lifingZefram2011-09-041-0/+1
|
* Add ext/re/re.pm to the @INC set for miniperl by lib/buildcustomize.plNicholas Clark2011-02-151-0/+1
| | | | | | | This avoids a build-time race condition where lib/re.pm might be read midway through the *second* copy of it (when ext/re/Makefile is being run). It also simplifies many [Mm]akefile* rules, which previously had a special case to copy it early.
* Use a buildcustomize.pl to set @INC in miniperl when building extensions.Nicholas Clark2011-02-151-0/+50
With the build tools now shipped in various subdirectories of cpan/ and dist/ we need to add several paths to @INC when invoking MakeMaker (etc) to build extensions. The previous approach of using $ENV{PERL5LIB} was fragile, because: a: It was hitting the length limit for %ENV variables on VMS b: It was running the risk of race conditions in a parallel build - ExtUtils::Makemaker "knows" to add -I../..lib, which puts lib at the *front* of @INC, but if one parallel process happens to copy a module into lib/ whilst another is searching for it, the second may get a partial read c: Overwriting $ENV{PERL5LIB} breaks any system where any of the installed build tools are actually implemented in Perl, if they are relying on $ENV{PERL5LIB} for setup This approach a: Doesn't have %ENV length limits b: Ensures that lib/ is last, so copy targets are always shadowing copy sources c: Only affects miniperl, and doesn't touch $ENV{PERL5LIB} Approaches that turned out to have fatal flaws: 1: Using $ENV{PERL5OPT} with a module fails because ExtUtils::MakeMaker searches for the build perl without setting lib, and treats the error caused by a failed -M as "not a valid perl 5 binary" 2: Refactoring ExtUtils::MakeMaker to *not* use -I for lib, and instead rely on $ENV{PERL5LIB} [which includes "../../lib"] fails because: some extensions have subdirectories, and on these EU::MM correctly uses -I../../../lib, where as $ENV{PERL5LIB} only has space for relative paths, and only with two levels. This approach actually takes advantage of ExtUtils::MakeMaker setting an -I option correct for the depth of directory being built.