[Libreoffice] merge issues ...

[Jan Holesovsky]

Hi Bjoern,

On 2011-05-31 at 14:30 +0200, Bjoern Michaelsen wrote:

> I just noted that commit
> bootstrap:f6ada6a7bf27fd8e0a50f0096604f541beca06e1
> got silently reverted by one of the recent merges. The commit
> itself is uncritical, but do we have a general problem here? It does
> not seem as if there was any manual intervention at all for this upon
> merge.

I investigated, and it happened because the code introducing the

$(info gb_LinkTarget_add_linktarget_objects,$(1),$(2))

was pushed to both master and libreoffice-3-4, but the code removing
this line just to master.  That correctly conflicted - and the person
doing the merge [me in this case - sorry for that] - picked the version
containing that line [either that I overlooked that, did not investigate
enough, or the tooling played tricks on me, and picked that
automatically - I don't remember].

> Silently dropping changes from master seems to be a serious issue to
> me.
 
All would be fine if either everything [the original code as well as the
fix] was pushed to both libreoffice-3-4 and master, or everything to
just one of these branches.

The conclusion is that whatever way anybody chooses (committing just to
release branch, or committing to master, and cherry-picking to the
release branch), the person has to be consistent in the choice when
there are subsequent fixes.

[And before you think "this wouldn't have happened if we never merged
the release branch back into master, because we would be always
cherry-picking" - I just have a counter-example of a fix that was
committed to master quite some time ago, but after the libreoffice-3-4
branch-off, and fell though the cracks - nobody noticed that it hasn't
been cherry-picked to libreoffice-3-4.]

Regards,
Kendy