Commit Graph

136 Commits

Author SHA1 Message Date
espie
d2fbe0a110 ouch, look for the actual location, not the package name, so that
we use built packages as much as possible (this is much faster than
regenerating packing-lists on the fly).
2007-06-03 11:57:36 +00:00
espie
009f44b53a allows saving/retrieving dump of objdump runs. 2007-06-03 11:01:10 +00:00
espie
1cc1790f05 add an option which simply dumps the stuff out. 2007-06-03 10:32:23 +00:00
espie
27af9a55a3 assorted cleaning-up:
- no longer any needed to put haslib and needlib together, since we
create it independently.
- replace the tests for opt_f by a distinct Recorder object, that either
keeps all binary names, or just some (bonus: this should allow us to
store the information the other way around).
- remove extra $db parameter which no longer servers any purpose.

Comment a few data structure.
2007-06-03 09:21:50 +00:00
espie
f10eb23d0b finish killing old stuff, you should have updated by now. 2007-06-02 12:23:28 +00:00
espie
3e81f47fb3 synch with pkg_add src/ changes 2007-05-23 10:34:15 +00:00
espie
9d2f3b66db old `check-lib-depends' has been dead for ages. Prepare for renaming. 2007-05-22 10:39:05 +00:00
espie
8eb23915f4 we don't bother building hardlinks either 2007-05-21 21:43:50 +00:00
espie
170d57a28f oops again: make sure we create the correct files. 2007-05-21 21:38:01 +00:00
espie
3876bf00e9 zap debug line 2007-05-21 21:24:56 +00:00
espie
49aaeb4f95 use the new code to ALSO report libraries recorded in WANTLIB (and used)
that are actually not reachable.

Reword the print-out slightly, I think it looks nicer ;-)
2007-05-21 21:23:45 +00:00
espie
e0e45027b6 oops 2007-05-21 21:02:14 +00:00
espie
3b31079021 unify reporting of lib issues 2007-05-21 20:52:33 +00:00
espie
023cfd6b44 Create error objects based on library stuff. 2007-05-21 20:39:27 +00:00
espie
ed3de34f07 separate usage of variables. 2007-05-21 20:20:50 +00:00
espie
440c45949a $o->method() becomes $o->method 2007-05-21 20:14:28 +00:00
espie
ee3c5a28f4 slightly clearer code: retrieve files through a FileSource, that abstracts
all details of archive extraction (or not).
2007-05-21 14:41:38 +00:00
espie
f438993d53 cut the main routine into scanning binaries/checking libraries. 2007-05-21 14:15:31 +00:00
espie
b61771ff87 generalize info files to zap auxiliary junk. Allows for manpages to change
without obvious negative consequences.
2007-05-20 17:27:51 +00:00
pvalchev
133a1b014c 4.1; beck 2007-05-14 16:16:17 +00:00
espie
c43a2cc6c2 use new-style plist visitors. 2007-05-13 08:03:47 +00:00
espie
2b5101e32f use newest PkgSpec interface 2007-05-12 14:56:21 +00:00
espie
a5d3f1822b shave a bit of time (roughly 5%) by noticing that package lists are
now more or less `unique', so we can avoid processing them more
than once.
2007-04-16 15:09:03 +00:00
espie
4d3e4bb38e shrink memory used by building only one single list per set of pkgnames.
Also use a simple `seen' cache for files only used by one package, since
they cannot participate in actual conflicts.

Shrinks memory for 250M to 190M, with no negative speed effects.
2007-04-16 14:36:25 +00:00
espie
8b9618e252 minor indent clean-up.
use auto-vivification to shorten code a bit.
pull out complicated code into its own function.
marginally faster than old version... some size decrease to do later...
2007-04-10 21:36:26 +00:00
espie
44415f04fb abort early if packing-list was not read correctly. 2007-04-08 12:36:04 +00:00
espie
2ad9b6ee63 error out in case something is not found, remove debug message. 2007-04-03 10:13:27 +00:00
espie
31721c0772 don't restrict -B to the case the packing-list comes from stdin. 2007-04-03 10:05:50 +00:00
espie
8016e70458 new mode: can work with packing-list on stdin, and a fake directory as -B
argument.
2007-03-31 15:26:45 +00:00
bernd
5183642076 Introduce -p option. This allows us to scan the ports tree for conflicts
without the need to have the packages around.

It's really easy now to check if a new package conflicts with other stuff
in the ports-tree:

$ find-all-conflicts -p /usr/ports newpkg.tgz

help & ok espie@
2007-02-19 19:50:50 +00:00
sturm
1cc933bba1 check build depends (currently duplicates most of check_run_deps())
ok bernd
2007-01-16 18:15:53 +00:00
bernd
64d1a3eb01 It's possible that the package version we get from the ports tree is
different than expected. Teach handle_dependency() how to cope with that.

Debugged with msf@ some time ago. Fix tested by msf@, sturm@ and steven@.

This also fixes the following error some people have seen:

"Can't use an undefined value as a HASH reference at
/usr/ports/infrastructure/package/check-newlib-depends line 234."

"go for it" espie@
2007-01-11 18:45:18 +00:00
espie
1ef1f2940b let check-newlib-depends use the new targets 2006-12-18 12:53:09 +00:00
bernd
12d8350649 Output beautification. One chunk from sturm@, two chunks by myself.
ok sturm@
2006-12-15 11:19:46 +00:00
sturm
e4f1e2c985 don't check build dependencies for now, these have to be compared against
indirect run dependencies which needs a little more code
2006-12-15 09:58:55 +00:00
sturm
0bcaa0f861 reduce the amount of direct dependencies by pointing out those which
are pulled in as indirect dependencies

ok espie@
2006-12-10 17:34:41 +00:00
espie
ea00df0a25 allows grabbing the packing-list from the ports tree.
suggestion from bernd
2006-11-11 13:11:45 +00:00
espie
321aced247 rename function for future change.
avoid re-reading the same plist for getting the same info (which doesn't
work anyways, so this avoids looking up a package for nothing).
let depwalk stash the pkgpath away for future use.
2006-11-11 12:44:26 +00:00
pvalchev
ea93d319d1 4.0 2006-10-26 00:41:45 +00:00
pvalchev
db68c59fd2 update list of architectures 2006-10-26 00:40:46 +00:00
espie
3291960295 oops, fix. 2006-10-23 11:19:35 +00:00
espie
872bae539e -f: display a full listing of which binary uses what library. Useful for
figuring out dependencies in a fine way for creating multi packages.
2006-10-23 10:54:45 +00:00
bernd
5e5fdfe2a4 Add libs which are reachable through dependencies to the cut&paste WANTLIB
output as well.

spotted by alek@, ok espie@
2006-10-20 18:10:25 +00:00
espie
7aa333e1ca output a simple set of WANTLIB lines you can cut&paste, by popular demand... 2006-10-20 15:09:24 +00:00
espie
992a0a28a0 sort information instead of putting it out in an haphazard way. 2006-10-20 14:40:29 +00:00
bernd
1db3e02224 Teach register-plist how to deal with the case where the recorded
plist in PLIST_DB misses the last line.

found & debugged by sturm@
ok sturm@, espie@
2006-10-19 00:15:46 +00:00
espie
55c0fd4b29 moved to ports/databases/sqlports 2006-10-15 09:42:17 +00:00
espie
dff6979a7a oops, forgot to commit FULLPKGNAME.
From Serge Basterot
2006-10-14 09:03:23 +00:00
espie
ccb088f096 let register-plist show stuff in a format ready for copy-paste.
Suggestion from alek@, took it one step further...
2006-10-02 18:48:01 +00:00
espie
6372292fc7 split things exactly like make does, avoid a few spurious error messages
from sqlite.
2006-09-19 09:57:58 +00:00