Re: CVS: commit dom abi/src/wp/impexp/xp ie_mailmerge.cpp

> commit-to-head-and-then-backport-trick-especially-so-close-for-releases
> > next time? Just to create the fake impression that
> > we do have a release
> > process in place... we've seen stuff break too often
> > already close for a
> > release :) Ofcourse it's not that we don't trust
> > your code Dom, because
> > you've never broken stuff like, say, link-grammar
> > :-P
>
> No. HEAD is in such a sorry state, that it's unusable
> for any of my needs. In fact, printing is so broken
> that I couldn't test this code in HEAD, and know that
> I didn't break printing...

Add DefaultGraphics="0x105" to your _custom_ scheme in the
AbiWord.profile. The pango graphics class is busted.

> So long as your backport
> policy is to blindly backport patches with the word
> "backport" in them without any testing, I fail to see
> how this pragmatically differs.

That's not true; if that's the case, then I shouldn't have any patches
in my backport queue left that are marked "Backport". Yet there are
still 'pending' patches which I didn't approve yet, because I know they
break stuff. Now except for some platform specific changes that I can't
always test and look innocent, I always:

1) see if I can reproduce the problem, of not, stop
2) apply the patch and see if it solves the problem
3) run the result through
   - my "diff" test for a set of documents, so see if any of the
     output changed as a result of the patch
   - run it through valgrind, before and after, and see if there where
     regressions
   - run it through gprof (should make this callgrind really), to see
    if there are performance regressions.

Now this process is not bulletproof, as I for example could not find any
errors in the patch from hub that improved spell check performance, yet
sum1 reported an error on win32. On the other hand, it catches most
errors, given the not-too-back-track-record for the last 20 or so
releases.

Mind that the breakage of 2.4.3 was not a result of any of my backport
actions, or lack of testing.

Bye!
  Marc
Received on Fri Apr 7 16:50:41 2006