[FFmpeg-cvslog] r14168 - trunk/libavformat/psxstr.c

[Måns Rullgård]

The Wanderer <inverseparadox at comcast.net> writes:

> M?ns Rullg?rd wrote:
>
>> Reimar D?ffinger <Reimar.Doeffinger at stud.uni-karlsruhe.de> writes:
>> 
>>> On Fri, Jul 11, 2008 at 04:41:44PM +0100, M?ns Rullg?rd wrote:
>
>>>> Just as I was thinking the previous commit message was bad...
>>> 
>>> I do not really mind that one, it is not (supposed to be) a
>>> functional change, so what "important" information is missing here?
>>> The affected file might not have hurt, but is seems a minor thing
>>> to me.
>> 
>> Yes, there is not supposed to be any functional change.  However,
>> bugs happen, and if I'm trying to figure out why something broke,
>> it's nice to be able to easily tell which commits might be relevant.
>> If the commit message mentions psxstr, I know I can safely ignore it
>> when debugging some other format, for instance.  Seeing only
>> "simplify", I have to look at the diff to determine the irrelevance
>> of this change.
>
> In what context will you see the commit message where the filename is
> not already present? The only places I know of offhand which have it are
> this mailing list and e.g. 'svn log', both of which already provide it
> (the latter implicitly, by way of the filename being specified on the
> command line).

git log --pretty=online, gitk, gitweb, ...

I'm frequently annoyed by the lack of detail in commit messages, and
I'm apparently not alone.  I'm afraid I can't give any scientific
evaluation of the usefulness of any particular log formats.

> Testing now to make sure I'm not being an idiot I do notice that 'svn
> log' can be run on a directory to get log results for that entire
> hierarchy, but I'm not sure I can think of cases in which it might be
> useful to actually do that.

I do such things on a daily basis.  One particular case I find useful,
is the git feature to display commits on one branch after another was
branch was created, or in other words, what you'd get if you were to
merge.

>> I'm tempted to add a hook enforcing at least three words in a commit
>> message.  It is simply not possible to be accurate enough in only one
>> or two words.
>
> I have a tendency to want to take statements like that as a challenge...

I count 14 words there.  Is that really the best you can do?

-- 
M?ns Rullg?rd
mans at mansr.com