Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
(devel/p5-PAR-Packer) Updated 1.052 to 1.059
1.059 2023-07-20 - correctly propagate the exit value of a packed executable Get the exit value from perl_destruct() instead of perl_run(). These may be different, e.g. if the value of $? is set in an END block. - clean up tests a bit - set up GitHub CI and add cpanfile (for CI only) - check more return codes in myldr/*.c 1.058 2023-06-12 - patch DynaLoader.pm to make intercepting DynaLoader::bootstrap work again for Perl >= 5.35 on Windows (see #74) $do_expand was changed in perl 5.36 from an (implicit) "local" variable to "my". This breaks how PAR::Heavy intercepts calls to DynaLoader::bootstrap: monkey-patching DynaLoader::bootstrap, DynaLoader::dl_findfile and DynaLoader::bootstrap while setting "local $DynaLoader::do_expand = 1". - reorganize modules etc used in tests - add test to run an XS module from a .par file - add a test for loading XS "glue" DLLs - remove unused code for packed DLLs ("pp --link ...") - nobody packs (non-XS) DLLs as embedded "FILE"s - nobody uses a $Config{version) sub directory when packing DLLs below "shlib/" in the zip 1.057 2022-11-29 - use a different method to mark executable built from "pp --clean ..." - scripts/par.pl: don't patch the string "__PASS_PAR_CLEAN__ \0" in the "boot" section of the executable and ... - myldr/boot.c: ... stop looking for the patched string - scripts/par.pl: add "\0CLEAN" in lieu of "\0CACHE" (and drop the 40-byte cache name below that) in the "trailer" section when generating a packed executable when META.yml indicates "--clean" was specified) - myldr/mktmpdir.c: allow "\0CLEAN" as an alternative to "\0CACHE" and set PAR_CLEAN=1 in that case - myldr/mktmpdir.c: implement find_par_magic() akin the one in script/par.pl - better CLT detection in MacOS (#70) [plk <Philip@kime.org.uk>] - use Getopt::ArgvFile with resolveEnvVars=1 suggested by Johan Vromans (@sciurius on GitHub) 1.056 2022-09-05 - Fix #66: patch myldr/boot for "pp --clean ..." without side effects - make sure myldr/boot contains exactly one string of the form "__PASS_PAR_CLEAN__ \0" so that there are no duplicates that may get split on chunk boundaries (myldr/boot_embedded_files.c) - simplify patching of this string (in $loader) to "__PASS_PAR_CLEAN__=1 \0" in script/par.pl - add a test for #66 (check for ephemeral vs persistent cache directory) - Revert "Fixes #62: rs6000_71 (AIX) "An offset in the .loader section header is too large."" PAR_CLEAN is set too late: at this point PAR_TEMP has already been set (and populated) to a persistent cache directory (/tmp/par-USER/cache-SHA1) instead of an ephemeral one (/tmp/par-USER/temp-PID). - Some code cleanup - replace some magic numbers with constants - use string interpolation (instead of concatenation) - clean up some convoluted C code 1.055 2022-07-03 - Fix #62: rs6000_71 (AIX) "An offset in the .loader section header is too large." Communicate pp option "--clean" to the generated executable in a different way. Previously this was done by patching "__ENV_PAR_..." strings **in the executable being built** and interpreting these strings in par.pl **at runtime**. Though hacky, this seemingly worked on any OS (without rendering the executable invalid). But the only information ever passed was the value of PAR_CLEAN and this was gleaned at build time from the value of JSONPath $par.clean from META.yml in the zip (set by pp by option "--clean"). Instead read and interpret "META.yml" in par.pl **at runtime**. - Fix: merge PR #58 from Philip@kime.org.uk: Adding support for running MacOS Universal binaries created via 'lipo' from already pp'ed and signed thin binaries - Make writing stuff more robust: check return value of print() and close() in some places. 1.054 2022-01-27 - Temporarily disable t/90-gh41.t (system("\\\\?\\some-absolute-path") fails) 1.053 2022-01-25 - Fix: merge PR #56 from from Philip Kime <Philip@kime.org.uk>: Updating OSX codesign fix util with fix for automatically added ad-hoc signatures on OSX 12+ - Fix: merge PR #42 from Andrew-Kulpa/master: Get exe size from file handle instead of path - Fix: prevent TABs in literal makefile fragments from being expanded - Cleanups: - par_findprog(): pass a copy of argument `path´ to strtok() - simplify searching for PAR magic - consolidate die()s - add a test for GitHub #41 - add error message when exec'ing the custom perl fails in myldr/boot - untangle %require_list and %ModuleCache
- Loading branch information