please test attempt at exclude option
Hallo, could you please take a look at fritsd/base-installer@3855f7eb
I wrote it to de-select systemd-sysv from the pkgdetails GETDEPS function.
It's probably not ready for production and has all its ugly DBG print statements still :-)
I still have to commit things for live-build
Dear Frits, first of all let me state your effort is becoming very useful to Devuan
Following up this preliminary discussion fritsd/base-installer@3855f7eb
I'm now reviewing your patch. Must say that the use of
strdupmakes me very nervous, so I'm trying to build test units that will facilitate debugging this patch and eventually also integration of future changes from upstream.
While I proceed with that, let me just congratulate with the achievement since a quick run on Devuan Jessie's
Packagesfile leads to success:
DEBOOTSTRAP_EXCLUDE_FIELD=systemd-sysv ./pkgdetails GETDEPS Packages init W: pkgdetails: skip init dependency on systemd-sysv I: pkgdetails: consider init dependency on sysvinit-core I: pkgdetails: used init dependency on sysvinit-core sysvinit-core
As you already said, Valgrind reports no memleaks on current HEAD, I see that too.
Running gcov (coverage) shows only some places where things could be optimized, i.e.
L312: if (*buf && buf[strlen(buf)-1] == '\ ') buf[strlen(buf)-1] = '\\0';
The value given by
strlen()could be pre-calculated considering this line is ran more than 300k times with the init dep example I'm using. Some similar things here and there and a DBG message leftover in run-deboostrap.c
Overall I'd have avoided so much dynamic allocation, but one can tell from your code that you have a very pragmatic and clean style of writing and it seems it all runs smooth.
As there is consensus among other developers, we can go ahead with this approach and perhaps even implement version parsing later on.
As of testing suite I'm now convinced is not really useful at this level, but we'll run more tests on using base-install and deboostrap both in the SDK and you of course for the fine work you are doing on live-build scripts.
After the last polishing of the package you may want to do I recommend we merge this patch in Devuan.Edited
I'm doing a bit of cleanup work based on your comments, expect 2 more commits soon.
building for unstable now. If it builds cleanly I'll add for ascii and jessie too.
unstable built succesfully. Building for ascii and jessie too.
All builds successful.
and thanks Daniel for driving the CI build, its reassuring to see you are in complete control of it, I still need to catch up with that.
Status changed to closed
Oh dear.... I forgot to bump the version - would have caused this. Will do that now and rebuild....
Oh, do you mean something like "debchange -i"? I'm trying to get a good understanding of the procedures.
I know that debuild chokes on changes that are not sanctioned by a changelog bump (and, rightfully so, I think, even though it can be very annoying)
Whenever we do a release/build the version needs to be bumped so that the new package is detected by the repository infrastructure. I use
git-dch -a -R -N <new-release version>and then commit that separately. (git-dch does have options for automatically committing that change too).Edited