On Sun, Jan 18, 2009 at 11:23:32AM -0500, microwaverich <richgrif@one.net> was heard to say:
Then I repeated the install of aptitude, and here's what it told me:
dpkg -i '///home/rich/downloads/aptitude_0.4.10-1_i386.deb' ;echo RESULT=$?
(Reading database ... 112547 files and directories currently installed.)
Preparing to replace aptitude 0.4.10-1 (using
..../aptitude_0.4.10-1_i386.deb) ...
Unpacking replacement aptitude ...
dpkg: dependency problems prevent configuration of aptitude:
aptitude depends on libapt-pkg-libc6.6-6-4.6; however:
Package libapt-pkg-libc6.6-6-4.6 is not installed.
Ouch. I forgot that the apt ABI had changed. So much for that clever
idea. I don't think we want to get into downgrading your apt.
But the one thing you did confirm is that apparently 0.4.11-2 doesn't
exhibit this bug. That gives us a starting point (and also limits how
far back we need to go, which will be a help). I think the version
range is narrow enough for "hg bisect" to be useful.
First, install the build dependencies of aptitude. You'll also want
mercurial.
# aptitude build-depends aptitude build-essential+ mercurial+
Now, get a copy of the aptitude source in the current directory:
$ hg clone http://hg.debian.org/hg/aptitude/head aptitude
$ cd aptitude
Tell "hg bisect" that the last known-good revision is 0.4.11:
$ hg bisect --good 0.4.11
$ hg bisect --bad tip
Mercurial will tell you that it's checked out a changeset for you to
test:
Testing changeset 1499:79935933f5ba (305 changesets remaining, ~8 tests)
136 files updated, 0 files merged, 8 files removed, 0 files unresolved
Now, compile aptitude by running
$ sh ./autogen.sh
$ ./configure --disable-werror
$ make -C src
When you're done, run ./src/aptitude as root:
# ./src/aptitude
If you can reproduce the bug, mark the current revision as bad:
$ hg bisect --bad
If not, mark it as good:
$ hg bisect --good
Then in theory you can just continue compiling aptitude (using the
steps above) and running it, telling "hg" at each step whether the
revision is good or bad. If you hit a revision that you can't compile,
you can try this:
$ hg bisect --skip
If that doesn't work, post the output of "hg identify" and the
compile error, and I'll see if I can figure out a workaround for you.
Otherwise, keep going (should take about 8 tries) until mercurial says
what it thinks the first bad revision is, then post the output here.
Thanks,
Daniel