2008-11-27 k * ready for the next release * for the first time we need the same trick we had already used in perlpatch2svn: write a APC-broken file out of the memory * current real state of my examples 2008-08-24 k * current state 2008-08-03 k * current state 2008-06-28 k * cleanup * quick hack to get an overview over unbaked patches 2008-04-29 k * trying to get rid of another file in one of the two patches * trying to get rid of two "reverse" patches * current state of (maintained) example files * it seems we have not tested for 5.9.0 correctly and it needs additional code 2008-03-22 k * update the axamples to my current practice * let binsearch always ask for --writedotpatch when calling buildaperl, really helps keeping track 2008-03-16 k * trivial smoker maintainance 2008-03-08 k * prepare a stable release; the eg directory sees new BBC breakpoints and an overhaul of the Makefile 2008-02-22 k * prepare a dev release * add proper descriptions to all tests * new potentially interesting breakpoint 2008-02-16 k * small doc fixes during a trainride; notes about how the 5.10.1 track should be supported 2008-02-12 k * blah 2008-02-06 k * sleep more between build sessions * stop promoting some very old perls because we rarely need them * notes about todos/bug fixes * two bugfixes at once: must set environment LANG to C to make sure we get the right messages from make; must verify if we have a file and return the relative path to it 2008-01-25 k * Property set for release after_284 * about to release 2008-01-23 k * Phillippe has established an "additional_tarballs/" directory 2008-01-20 k * Property set for release after_281 * missing PREREQ_PM on version.pm * signature stuff * * Changes - doc me * Makefile.PL - new dependency on Module::CoreList * README - adjust to the new prerequisites with a link to README.additional_tarballs * README.additional_tarballs - new file describing the manual tweaks necessary to get the support for additional tarballs working * Releasenotes - removed * eg/Makefile - get rid of the copy-links option and suggest better rsync strategy * lib/Perl/Repository/APC/BAP.pm - support additional tarballs - rename variables for better self-documentation - adjust docs * lib/Perl/Repository/APC.pm - add metadata about all 15 tarballs that were not supported by the APC repository - add support for the 15 - adjust the internal datastructure in a backwards incompatible way (hence the version jump) and adjust all methods to understand it - new routine get_diff_dir() because now diff directory and "next" perl can be different * scripts/buildaperl - harvest the 5th return value of P:R:BAP::translate to find additional tarballs - rename a few variables to be easier to understand * scripts/apc-overview - adjust to the new internal structures of the P:R:APC module - improve Usage messages * scripts/patchaperlup - reduce the noise produced during normal use * t/bap.t * t/apc.t - improve the tests and rename variables to make them better readable; add a test for 5.9 and patchlevel 30000 2008-01-19 k * stealing cosmetic changes from the CPAN.pm repo * taking notes about the current state before; small doc fixes 2008-01-03 k * Property set for release after_275 * prepare release 2007-11-11 k * Property set for release after_273 * prepare a release * fix missing commas * more exceptions from the cleanup routine * tiny simplification in the manpage 2007-11-10 k * check the return value of the prep script and retry to run it if it fails (3x) * be more verbose so we can spot the bug next time 2007-11-02 k * comment on looping * make sure that prep script runs after a buildnext() run * unneeded whitespace change that just happened * exempt plenty of perls from being removed by thinnerrepo * publish the repo in META.yml * sorted CONFIG options; disable trimtrees because it is inherently dangerous 2007-09-09 k * stop linking empty files * start using relocatableinc 2007-08-12 k * Property set for release after_258 * preparing a release * create the .patch later otherwise it will be deleted * buildaperl: implement writedotpatch; patchaperlup: cosmetics * patchaperlup: implement writedotpatch * re-indent * move subroutines together and write a large block around MAIN; no functional change 2007-07-22 k * make the most important pairs of patch neighbors undeleteable 2007-07-10 k * Property set for release after_250 * prepare new release * applied a patch by Father Chrysotstomos to better integrate with gzcat which is often available on OSX * do not delete very young directories, somebody might still need them right now * improve disgnostice * personally I need to reduce the number of perls on stick from 1024 to 768 because they get bigger with 1260 modules 2007-05-02 k * blah 2007-04-24 k * Property set for release after_243 * prepare a new release 2007-03-07 k * blah 2007-02-18 k * blah 2007-02-14 k * Property set for release after_239 * fix a badly constructed test 2007-02-12 k * Property set for release after_237 * add Changes * prereq Digest::MD5 added 2007-02-11 k * Property set for release after_234 * s/Releasenotes/Changes/ 2007-02-10 k * adjust wording to my current taste; disable the 99999 hack when --exact-bounds is given * fix binsearch to take the highest perl we have if the upper bound is even higher (allow for 99999) * remove an unneeded concurrency blocker (which probably was not even working correctly) * chmod the temporary directory in the cache to 0755 * remove the code that stops to ask for confirmation just because it is a testfile being removed 2007-01-25 k * blah 2007-01-02 k * blah 2006-12-01 k * the semi had to go for a && 2006-11-30 k * for the first time we have a makefile that builds bleadperl AND a CPAN megainstall with one make command * bug workaround for a few hours disabled again because the cause is fixed * keep output in a file with timestamp and add the -b switch to rsync so we can always see what exactly has changed when a patch gets remastered * prepare a change if we need to change the -p swich occasionally 2006-10-22 k * improve the coverage * improve kwalitee by adding license stuff 2006-10-20 k * no default for PERL_REPOSITORY_APC_REPO anymore, now distroprefs is used to set the value * in trimtrees we were always testing with a wrong perl 2006-09-28 k * blah 2006-09-10 k * Property set for release after_215 2006-09-09 k * patch 28791 introduced a zero length file and we knoe no way to create it 2006-08-19 k * blah 2006-04-14 k * blah 2006-02-25 k * Property set for release after_211 * MM version * make the Makefile.PL more similar to the one in CPAN.pm: I just made a distro with old SIG -- AGAIN! * Property set for release after_208 * copying MANIFEST seems nosense because it is a link??? * Property set for release after_206 * copying META.yml seems nosense because it is a link??? * Property set for release after_204 * copying SIGNATURE seems nosense because it is a link??? * Property set for release after_202 * make svn deprecation more visible; add version testing to t/script; do not test the svn scripts anymore; add the version option to apc-overview * makethe deprecation mor visible * trying to skip over sv.h#252 in everything but in 27297 * bleadperl 27302 was a straight reverse and -N was inappropriate * make the _arg variables look alike * reveal more switched to patchlevel.h 2006-02-05 k * we did the CP stuff twice, in distdir and release; the former should do * Property set for release after_194 * SUPER without package does not work * MY:: and a comment why the distdir target is there * Property set for release after_191 * ahh, I missed the SIGN => 1 stuff in the Makefile.PL! * Property set for release after_189 * remove the sign target from the dist target; adding distdir stuff from CPAN.pm without recognizing what exactly it does: the last release contained a wrong signature and I need to make sure that this does not happen again * Property set for release after_187 * add a signature test 2006-02-04 k * Property set for release after_185 * trial and error whats going on with make release * Property set for release after_183 * trial and error whats going on with make release * Property set for release after_181 * trial and error whats going on with make release; adding eg/relocate-perl.pl * Property set for release after_179 * how crazy to find a bug immediately after uploading * Property set for release after_177 * typo in release target * Property set for release after_175 * introduce the release target; fix the script.t test to accept --help as an alternative to --h * Property set for release after_173 * add the cachefilter option to binsearchaperl; add entry to the FAQ about Module::CoreList; write a nicer ChangeLog file 2006-01-02 k * simple script to let an installed perl live in a different directory 2005-12-03 k * blah 2005-11-30 k * Property set for release after_169 * forgot t/pod.t in the manifest * Property set for release after_167 * MANIFEST, Releasenotes: Add Releasenotes eg/trimtrees.pl: Add workaround bug in File::Compare * add a Test::Pod test 2005-11-18 k * now he dies are also dumpvalued * count $usedspc correctly when a file is not readable, extend tests to control the effect at least viaually * counting also files we cannot read * improve error messages with dumpvalue against strange filenames and more $!; spotted a missing close; added tests against the error messages except for the compare message because I cannot produce it (yet) 2005-11-17 k * one nice senence fr the docs * extend the maxlinks semantics to let us prepare a linktree to a subsequent cp -al 2005-11-16 k * add getopt to the testscript so we an use it to test maxlinks 2005-11-12 k * trimtrees.t ready for release as testscript * maxversion supported via --maxlinks, defaults to off; bug fixed in maxlinks in that only the number of links on $first is relevant, not the sum of links on $first plus number of links on $cand; replace several returns with gotos because return misses the xreport * Functionally equivalent to Erik's version but with minimal difference to my revision 154 and with a simple option to turn maxversion support off * Eriks version as of 2005-11-10 2005-11-10 k * simplify with File::Temp but still no proper test * factor out xreport * off by one error * just an idea for a test script, not yet good for distributing * shake out the bug with "too many links" error condition; this the version I sent to Erik Schnetter * appplied patch by Eric Schnetter that prevents a failure on ReiserFS when some limit on the number of hard links is reached (2^16 about) 2005-08-30 k * blah 2005-08-09 k * Property set for release after_146 * now it was my bug to forget a -v * I think I found the culprit, I believe that svn output has changed * there must have gone something wrong between 132 and now because APC.pm does not have the cpan:release property set higher than 131 * except that a German locale also introduces German strings, sorry * without locale setting svn still writes weird log output * now the ChangeLog was sorted from 1 ascending * still missed how to get a correct ChangeLog * fighting against setversion * chlog target did not reflect the really very latest checkin, only up to the version of the current directoy * update for a more recent svn version * exempt symlinks from trimtrees.pl; patch courtesy Marko Mäkelä 2005-06-23 k * blah 2005-04-02 k * nothing spectacular, just what cumulated over time while I forgot to checkin 2004-04-19 k * Warning more explicit/educative when patchls not found The eg/Makefile did not rsync old stuff 2003-09-14 k * Property set for release after_131 * * binsearchaperl: also needed a special case for 5.6.2 * rm-rf-allbutlatest.pl: was only keeping one maint directory, now keeps one per perl-base-version * eg/Makefile: forgot to rsync the 5.6.2 sources * BUGS: explain the situation for the record perlpatch2svn: fix the special case 5.6.2 apc2svn: ditto eg/Makefile: start building 5.6.2 to see how it goes * perlpatch2svn: cosmetics as preparation * Wrote short explanation of the workaround in rev 124 to the BUGS file * Found some duct tape to let the 5.6.2/3 branching problem disappear This is a temporary fix, it cries for a solution within the APC. BAP.pm: special-case the "maint-5.6/perl-5.6.2" branch APC.pm: identify the maint-5.6/perl-5.6.2 branch apc-overview: new printf to align the output eg/Makefile: my personal prefs changed again 2003-09-08 k * binsearchaperl: downgrade a die to a warn for the case we find an installed-perl/... directory without a perl binary apc2svn: doc patches eg/Makefile: introduce the -Dnoextensions option to safe time 2003-09-06 k * Property set for release after_121 * perlpatch2svn: cosmetics * Clean up all the debugging noise and add a --debug switch to apc2svn. Remove the "no svn test" bug from the BUGS file. Clean up svn.t as well. Ready for a release with the directory removal bug in apc2svn fixed. 2003-09-05 k * vn.t: just commenting and disabling the test again * svn.t: tweak until we see the bug reproduced APC2SVN.pm: get rid of double negation perlpatch2svn: make open3 usage & debugging cleaner 2003-09-04 k * svn.t: make it useable but disable it for all users (too slow) buildaperl: fix an updir bug, rename variale "this" to "dir" perlpatch2svn: comment extensively the bug we were trying to fix * svn.t: just comments APC2SVN.pm: introduce global variable $DEBUG perlpatch2svn: make the svn up conditional, not mandatory * bap.t: cosmetic change svn.t: small progress, still not useful APC2SVN.pm: more debug info output which needs to be eliminated later APC.pm: silence an unini warning perlpatch2svn: insert a mandatory "svn up" to see if that fixes all problems 2003-09-03 k * svn.t: skeleton of a new test which does nothing yet MANIFEST: add svn.t APC2SVN.pm: more debugging output * BUGS: talk about missing tests APC2SVN.pm: the now added deletes needed to be sorted perlpatch2svn: add some "mysqystem" noise to make debugging easier apc2svn: add locatime to mysystem output FAQ: harmless tweak 2003-09-02 k * * APC2SVN.pm: -- switch back to the unless or not style because the other variant misbehaved -- glob yourself, do not let the shell do it, because the shell globbing will transmit a star if there is no match -- delete_empty_dirs now also returns the names of the directories so that the caller can add them to the targets array * perlpatch2svn: -- add the return value of delete_empty_dirs to the @targets array but filter such that no repetitions occur * APC2SVN.pm: cleanup before debugging perlpatch2svn: cleanup before debugging Both get more labels, shorthands for complicated logic, short lines * APC2SVN.pm: fixing a case of unini warning * Add a BUGS file and confess that 5.6.2 doesn't work * Makefile.PL: add PREREQ_PM buildaperl: fail a bit more verbosely (might need even more) apc2svn: sort the options, add option bounds FAQ: add a note about portability eg/Makefile: add a note about --copy-links 2003-08-15 k * Property set for release after_106 2003-08-14 k * podifying * Code reformatting, remove unused code * User feedback more sensible * better selfdescribing variable names * Fix the extreme case where the cache is on multiple partitions * tempdir didn't cleanup completely 2003-08-13 k * Property set for release after_99 * Fix a race condition so that unlimited instances of buildaperl can run simultaneously. 2003-08-11 k * Property set for release after_97 * It's not easy to do it in the right sequence: META.yml, SIG, ChLog * Property set for release after_95 * MANIFEST.SKIP didn't allow the other distribution here * Property set for release after_93 * cp -v into DISTVNAME doesn't work anymore, it is a link now * Property set for release after_91 * My ChangeLog was not in MANIFEST, cpansign complains * Property set for release after_89 * bap.t: Test was out of sync with reality and a bug was hiding this for quite a while apc.t: Change the default path to the repository MANIFEST: add META.yml and SIGNATURE and all that BAP.pm: documentation was also not reflecting reality Makefile.PL: support META.yml and SIGNATURE MANIFEST.SKIP: up to reality * only keep 50 perls in cache * cleanup the grown eg directory * add some small tools to the eg directory 2003-08-03 k * Allow perl-0@1 as it is the beginning of the repository, first patch * buildaperl: make sure that $lev is > 0 if it is given as argument * Parse the output of make and work around the problem that old perls have with new gccs 2003-07-16 k * better reporting through $0 new option --die-on-error 2003-05-01 k * Property set for release after_80 * * binsearchaperl -- fix --prep switch to also act for the middle perl. * fix FAQ where the --prep switch needs to be mentioned. * * BAP.pm -- rename variable $last to $prev -- improve error message "patch $lev is not part..." to show neighbors * Makefile.PL -- Do not use svnversion for a single file (not supported anymore from subversion) * binsearchaperl -- introduce --prep switch to e.g. install modules * small FAQ tweaking * eg/Makefile now works with --copy-links 2003-03-13 k * Property set for release after_77 * Makefile.PL: workaround broken svnversion * Property set for release after_75 * Property set for release after_74 * apc2svn: add an argument check to myls 2003-03-12 k * APC2SVN.pm: make a warning look less severe * perlpatch2svn: tiny doc patch * perlpatch2svn: whitespace only patch; fix indenting * * perlpatch2svn: -- include branch in $0 * * apc2svn: -- usage fix File::Spec * * perlpatch2svn: -- fixed a bug in filtering out patches from wrong branches -- support for skipping unneeded patches -- Assign to $0 the current status -- use File::Temp for TMPLOGFILE * apc2svn: -- check for makepatch version > 2.0 -- use File::Temp for temporary files and remove them when not needed anymore 2003-03-09 k * Property set for release after_66 * * FAQ: --broke the FAQ out of the binsearchaperl manpage and made it a standalown document. * lots of small changes to the documentation * all scripts now have the --h switch which always exits with status 0 2003-03-08 k * Property set for release after_64 * * eg/Makefile -- rewritten to only connect once * binsearchaperl -- add to the FAQ question about incomplete mirror of repository * perlpatch2svn: -- patch from Rafael: add encoding=latin1 to the commit parameters. * SVN2APC.pm: -- prototype all functions, C< dir_will_be_empty > was called as a method! * apc2svn: -- die on more system() fails. We were too tolerant, we may still be too tolerant. 2003-03-02 k * Property set for release after_62 * * binsearchaperl: -- argument validation on --branch -- allow --bounds lower-upper with lower==upper, useful for --show-cache -- fixed a few off-by-one errors in the --bounds logic -- exit instead of die when maxbuild is reached -- added a FAQ section to the manpage; first FAQ is about handling modules 2003-03-01 k * * buildaperl: -- assign to $0 as soon as branch and level are known -- ./Configure cannot be run when execute bit isn't set. Let's call sh Configure instead. -- and use './perl installperl' instead of './installperl' -- shorten too long lines Makefile.PL: -- Switch from ${} to $() for nmake -- File::Spec'd eg/Makefile: -- now the default target does nothing to protect the innocent * Gross File::Spec'ing of several scripts and modules, but given up as I cannot test anywhere but on Linux * patchaperlup -- "--start 0" didn't understand that the argument was actually 0. Checking for definedness now instead of truth. * binsearchaperl: -- didn't recognize perl-0@60 as a perl directory in installed-perls -- made formatting of final message in case of identical results more readable 2003-02-27 k * * APC.pm: -- confess instead of die for better debugging * binsearchaperl: -- remove unused C -- more helpful error messages if installed-perls is still sparse -- buildnext now checks if $next is valid and helps user if not -- more helpful messages wrt %NOSUCCESS -- move informational messages to STDERR, only expected things go to STDOUT -- switch to trying random middle points when we have encountered "no success": it's too likely the neighbors won't compile either -- assign to $0 where we are currently search * buildaperl: -- relative pathnames in prefix need to be composed to absolute ones because we are in the wrong directory when using the argument. * Makefile.PL: -- fix an "uninitialized" warning 2003-02-26 k * Property set for release after_58 * * README: -- rewrite in 2-column-mode. * * patchaperlup -- adjust the format of warnings to be consistent -- do not try Changes at all if --start option is given 2003-02-25 k * Property set for release after_55 * Makefile.PL: more documentation * Property set for release after_53 * Makefile.PL: add documentation of the svn tricks * Property set for release after_51 * forgot MANIFEST.SKIP in MANIFEST * * Added MANIFEST.SKIP * Makefile.PL -- new tricks with svnversion to keep APC.pm at the leading edge * all *.pm files have again $Rev tricks to determine $VERSION * Testing svn rules in Makefile.PL * Makefile.PL: add Rafael's URL * * binsearchaperl: -- new option --show-cache * Makefile.PL: -- more appropriate die message when --setversion is active * apc2svn: -- trim the output * Makefile.PL: die later if version_diff * APC2SVN.pm: unmatched =back * buildaperl: tweaking the Info line * * all *.pm: -- hardcode the version number and let it be written by the "setversion" rule in the Makefile * APC2SVN.pm: -- import dirname. This was lost on the transition to APC2SVN -- remove some overly verbose output * Makefile.PL: -- remove the crufty verifyversion in favor of setversion: Letting subversion set the version number is not very CPAN-friendly. People may not notice updated versions because we only changed a script. As a bonus we get VERSION_FROM back. * buildaperl: -- patchlevel.h above 18749 now used for adding a comment * apc2svn: -- trim verbosity -- the '#line 4' trick to have always correct line numbers 2003-02-24 k * * BAP.pm: -- also return the first patch so that buildaperl can construct the --start option binsearchaperl: -- left-trim the long status line that echos the path to each perl buildaperl: -- always construct the --start option patchaperlup -- mention ccache in the documentation README: -- do not discourage apc2svn that strongly 2003-02-20 k * * added APC2SVN.pm, perlpatch2svn, and apc2svn * * added APC2SVN.pm, perlpatch2svn, and apc2svn * Makefile.PL: fixing verifyversion * Makefile.PL: verifyversion before dist 2003-02-18 k * * APC.pm -- bug in closest() when $wanted argument is 0 * apc.t -- test for the bug above -- use $i instead of hardcoded values. No I don't want Test::More in this package. Yet. * * Makefile.PL -- ChangeLog generation, take 4 * * Makefile.PL -- ChangeLog generation, take 3 * * Makefile.PL -- Spaces->Tabs * * Makefile.PL -- create the ChangeLog from Subversion 2003-02-17 k * * new test t/scripts.t * binsearchaperl -- docs about the installed-perls/ directory and the danger of mixed config options * buildaperl -- escaping @ in string * scripts/apc-overview -- add a header line to the output * added apc-overview to MANIFEST and EXE_FILES * * APC.pm -- added apcdirs method so that external scripts need not access object data directly * buildaperl -- small doc fixes * adding apc-overview * eg/test_full_apc.pl -- use apcdirs() * README -- rewording 2003-02-16 k * * buildaperl -- when renaming perl-5... to perl-p-5... I missed the early check "Directory exists, giving up". No reinstantiated -- Remove the code that changed patchlevel.h forever -- Removed the "tee" and the whole patchaperlup.out. As we check beforehand if the argument is correct, we do not have to read the output at all. * APC.pm -- pseudo checking * pseudo checkin to get a new VERSION in APM * * buildaperl -- branches were skipping many patches. I wonder why we skipped many but not all. Danger, Will Robinson! * * BAP.pm -- 5.6.0@7424 in branch maint-5.6 was rejected because we did not store the initial value for @ver in $last. * bap.t -- test for the fix above * * bap.t -- better skip message * * bap.t -- protect against inexistent repo * * Makefile.PL -- determine the VERSION from the highest included $Id * * buildaperl -- delete wince/perl.ico from MANIFEST if it doesn't exist * eg/Makefile -- updated to my needs * * buildaperl -- delete wince/perl.ico from MANIFEST if it doesn't exist * eg/Makefile -- updated to my needs * * APC.pm: -- documented closest in a comment, not in the manpage * buildaperl -- now accepts a formidable range of arguments using BAP.pm -- renames to a new directory name that indicates trunk/maint as in perl-m-5.8.0@18701/ vs. perl-p-5.8.0@18718/ * eg/Makefile * added P:R:APC:BAP to parse and translate the argument to buildaperl * added a test bap.t for APC/BAP.pm * t/apc.t -- now accept envariable to set repository 2003-02-15 k * * APC.pm: -- more input checks * all scripts get a "# line 4" directive at line 4 because the installed versions get 3 lines of shebang magic and then everything is off by something * buildaperl -- support argument "@" alone to pick latest patch of latest version of a given branch * move test_full_apc.pl to the eg/ directory * add my Makefile to the eg directory * add a README to the eg directory * buildaperl: bugfix * buildaperl: bugfix * buildaperl: forgot to remove debugging code * buildaperl: could not reach last version of a branch 2003-02-14 k * * added a readme * * APC.pm: -- $VERSION again * * APC.pm: -- $VERSION -- more input checking on patches() * buildaperl: -- use {first,next}_in_branch to determine what we have to do. The old code was optimized for perl branch and did not work for 5.005_01 -- remove nocompile and introduce a better noconfigure instead -- disable "patchlevelscript.3.txt" patching: this is private -- --remo was broken, a chdir was missing * patchaperlup -- many entries for %SKIP * test_full_apc.pl -- ad hoc changes for *really* running it * * APC.pm: -- confess if get_{to,from}_version are called without $patch * add binsearchaperl * buildaperl: -- remove prerequisites from manpage * * APC.pm: -- confess if get_{to,from}_version are called without $patch * add binsearchaperl * buildaperl: -- remove prerequisites from manpage * * APC.pm: -- did not even compile * testing * propset * * APC.pm: -- did not even compile * * added buildaperl from a different repository * started test_full_apc.pl * APC.pm: -- added method tarball 2003-02-13 k * -- patch_range and version_range now without code duplication -- in last checkin forgot to mention closest() which remains undocumented. Seems internal to me. * -- added patch_range and version_range: yes, I could have made them with less code duplication. It was my original intent, but then I forgot. * populate