summaryrefslogtreecommitdiff
path: root/mk/bulk
AgeCommit message (Collapse)AuthorFilesLines
2007-07-14Add support for the upcoming pkg_install-20070714 which now includesadrianp3-14/+53
audit-packages.
2007-07-02Remove USE_MTREE support. Keep one copy (the NetBSD version) forjoerg1-5/+5
the purpose of print-PLIST in plist/common-dirs.mtree. Discussed with wiz@, no objections on tech-pkg@.
2007-06-15Fix PR 34139 for bulk builds too.wiz1-12/+12
Patch from Sergey Svishchev.
2007-06-15The option --prepare only generates the bulk build database.rillig1-1/+16
2007-05-31If a package is already installed, bulk-install should not try to buildrillig1-2/+3
it again.
2007-05-23When removing an already installed package, the correct package versionrillig1-3/+4
is reported. Fixed the output by not nestedly redirecting the output of "bmake debug".
2007-05-20Add all packages that are used by pkgsrc/bootstrap to the list ofkristerw1-2/+4
packages that should not be built on non-NetBSD systems.
2007-05-19Revert previous (1.140), committed by accident.wiz1-5/+2
2007-05-19On Linux (at least Red Hat Enterprise 3), openssl needs librarieswiz2-5/+11
from /usr/kerberos and /dev/random in sandbox. On FreeBSD, use devfs to populate /dev in sandbox. From Sergey Svishchev.
2007-04-15It makes more sense to run "bmake debug" when the WRKDIR is stillrillig1-2/+2
populated with all sorts of files, since they provide the most useful information.
2007-03-22The egrep line can be removed when audit-packages becomes fast enough.rillig1-1/+3
2007-03-07A hint to the real problem is a nice addition to an error message.rillig1-2/+4
2007-03-07Added the BULK_ID to BULK_BUILD_ID_FILE, so that multiple bulk builds canrillig1-2/+2
share the same pkgsrc tree in read-write mode. Fixes PR 35939.
2007-03-06The first few lines of a bulk build log don't need to be printed on therillig1-9/+6
terminal. The wrapper warnings are included in the build log, no matter if it succeeds or not.
2007-03-06Removed the unused "arch" variable.rillig2-8/+4
2007-03-06The warnings from the wrapper log are added to the report file.rillig1-1/+8
2007-02-23The post-build-only option also needs to load the variables.rillig1-1/+2
2007-02-21At the very beginning, the variables are extracted by running bmake inrillig1-54/+24
the pkglint directory. This part had been more time-consuming than necessary. Now it only needs one call to bmake instead of n.
2007-02-21- also find possible listings in {whatever}foo variationsjschauma1-2/+2
- use PKGVULNDIR from env for correct location of pkg-vulnerabilities
2007-02-20Only invoke audit-packages if we do find the package-name prefix injschauma1-6/+8
the vulnerabilities file. This cuts down significantly on the time spent running sort-packages: On 3228 packages before: 20259.43s real 9888.82s user 13138.60s system With this change: 6595.97s real 3215.58s user 4265.13s system
2007-02-19The error output of "make debug" may be useful to know, too.rillig1-2/+2
2007-01-30Remove some legacy variable handling after 2-3 branches.wiz1-16/+1
2007-01-28Remove old .bulk-not_available (NOT_AVAILABLE_FILE) files that are leftkristerw1-1/+7
from previous run.
2007-01-24Empty files are so ambiguous, so better leave a friendly message in therillig1-2/+2
STARTFILE.
2007-01-24Quick fix: The main code of the bulk build cannot be run in the shell'srillig1-2/+2
"set -e" mode, since it would fill up the temporary directory. Reverted that change.
2007-01-22A big cleanup:rillig1-66/+40
- The "header" of the build log is always printed, not only when PRECLEAN is set. - The variable "installed" was redundant. Removing it saves 5 lines of code. - When the package is outdated and currently installed, it suffices to just run pkg_delete. The "make deinstall" is too expensive and doesn't do anything further. - Completely rewrote the section that deinstalls unneeded packages. It seems to me that the pkg_info -Q flag has been introduced after the bulk build was initially written, so it wasn't used here. - A package that doesn't get deinstalled with a simple pkg_delete probably has a reason for that. Don't try to run pkg_delete -f after it. - To remove all packages, it isn't necessary to iterate over the installed packages. Just pkg_delete -r "*". - Replaced "required package" with "dependency" in the log messages, since that is shorter. Maybe the term "required package" comes from the time where packages also depended on shared library versions, so that terminology made sense back in 2000. Anyway, that's long ago.
2007-01-21Fixed misleading indentation.rillig1-3/+3
2007-01-18pkgsrc bulk builds need much CPU time, but do they also need much stack?rillig1-3/+4
2007-01-15Made the database lookup for deinstalling the unneeded packages lessrillig1-10/+13
obscure by using awk instead of sed.
2007-01-08Packages that are not available on the current platform (and therillig2-13/+36
packages depending on them) are listed in a separate category in the bulk build report.
2007-01-08Shortened the code.rillig1-4/+2
2006-12-23Bump date for previous.wiz1-2/+2
2006-12-15Remove trailing tabs.martti5-10/+10
2006-12-15Remove trailing spaces.martti3-8/+8
2006-12-14Use '-o' instead of '-or' in find(1) arguments as the former seems todmcmahill1-2/+2
be more portable (works on at least NetBSD, SunOS-5.9, and Linux). Addressed PR pkg/35245
2006-12-14If a package build fails, append the output of 'make debug' to the logdmcmahill1-2/+5
since sometimes there are questions about various variables and settings. This should help in people trying to investigate build failures.
2006-12-14add /usr/sadm to the list of sandboxMountDirs for SunOSdmcmahill1-2/+2
2006-12-13include the PKGSRC_COMPILER setting in the reportdmcmahill1-1/+3
2006-12-10Fixed the creating of numerous zombie processes, which is a bug in therillig1-1/+8
NetBSD /bin/sh, reported in PR 35227.
2006-12-09report correct pkgcount by first cd'ing to PKGDB_DIR, then FINDing in there.jschauma1-2/+2
(else -prune will cause us to return immediately, always reporting 1 installed package even if we have more)
2006-12-09straighten out FIND command to figure out if a package needs to be rebuilt.jschauma1-2/+2
This should fix the problem reported by wiz@.
2006-12-07Due to a misunderstanding of mine, preventing to upload restrictedrillig1-2/+9
binary packages has not worked in the last few months. Now I fixed the rsync command line to use --files-from instead of --include-from, which does the right thing. Tested in a directory of about 20 binary packages, including one that is restricted. Everything worked as expected. Fixes PR 35199.
2006-11-27- modularize functionality into a number of distinct functionsjschauma1-301/+357
- add the option "-p|--post-build" to allow you to only run the post-build portion of the bulk-build
2006-11-26The bulk builds now use the new target can-be-built-here, which checksrillig2-3/+24
if some of PKG_SKIP_REASON, PKG_FAIL_REASON, NOT_FOR_*, ONLY_FOR_* are set. The code uses some shell trickery to avoid the use of subshells, since these would prevent the variable assignments from being effective.
2006-11-26- Run a single find(1) command to get the list of new files rather than runningjschauma1-4/+4
two finds+two greps. - Run a single find(1) command rather than ls | grep to determine pkg count.
2006-11-19nuke two Useless Use of Wc and let awk do the legwork. This saves usjschauma1-6/+8
two forks per package.
2006-11-06Also add bin/nawk to the tarred-up files.tv1-3/+4
2006-11-06Add lang/nawk and sysutils/install-sh to the "don't blow away" packagetv1-3/+3
list for bulk builds. (Revert previous on bootstrap; the PKG_PRESERVE hack didn't work as expected.)
2006-10-13Creating an archive from the working directory may take some time forrillig1-1/+2
big packages, so print an informational message before doing that.
2006-10-07Renamed the "pack" option for _PRESERVE_WRKSRC to "pack-if-failed",rillig1-4/+5
since that's what I really wanted. It's just one day old, so I hope that nobody has gotten used to it.
generated by cgit v1.2.3 (git 2.39.1) at 2024-09-23 09:18:32 +0000