Re: [Bug-wget] [Bug-Wget] Policy on commit messages

[Darshit Shah]

On 11/01, Tim Rühsen wrote:
> Am Samstag, 1. November 2014, 05:43:11 schrieb Darshit Shah:
> > Hi,
> >
> > In the last few days we've seen a huge flurry of activity and a large number
> > of patches have been committed to the code base. As someone who was
> > traveling and hence unable to keep up with the mailing lists, I tried to
> > catch up by looking at the code base and the changes that have been made.
> >
> > It would turn out that the commit messages aren't very informative on the
> > last few commits. Unless I know the context of the change from the mailing
> > list, it would be very difficult to understand why that particular change
> > was important. Both the ChangeLog and commit messages usually only describe
> > the change in literal terms but not in their logical reasoning.
> >
> > I propose that henceforth, all commits be reviewed for detailed descriptions
> > of not only the syntactical changes, but also the logical reasons behind
> > the change. This will help us in the long term to follow through the
> > changes in the code base and also make it easier for new contributors to
> > understand why some things are the way they are.
>
> As I understood, for GNU projects we have very detailed ChangeLog entries and
> thus very short commit messages.
The problem with ChangeLogs is that they only specify the changes as a fact. Not 
the actual logic / reasons behind them which is what we'd like to track in good 
commit messages.

> I am not going to make detailed ChangeLog entries AND detailed commit
> messages.

Sure, I personally find it too tedious to write both a ChangeLog and a commit 
message.
> If we move to detailed commit messages (I would appreciate it), then we should
> autogenerate ChangeLog files from the commit messages. You will find many
> projects out there already doing so.

I am all for autogenerating the ChangeLogs.
> If we can agree upon a frame or method, I would also appreciate a 'reason
> field' for the changes. If it is not redundant. E.g. a commit message like
> 'fixing double free()' stands for itself.

Precisely what I'm looking for!
> Tim


--- end quoted text ---

--
Thanking You,
Darshit Shah

pgpfE64LMV_Z9.pgp
Description: PGP signature