On Nov 17 Mark Brown wrote:
On Wed, Nov 17, 2010 at 01:53:35AM +0100, Jiri Kosina wrote:
On Wed, 17 Nov 2010, Stefan Richter wrote:
Why should we codify our conventions in MAINTAINERS to accommodate the specific problem of virtually a _single_ patch author?
It seems to be the way we're heading in general - look at all the recent work on MAINTAINERS and get_maintainer.pl. There seems to be a genral push to make all this stuff automatable.
get_maintainer.pl, used with judgment and together with "gitk the/patched/source.c" is nice not only for people like Joe who regularly work tree-wide but also for ones like me who only rarely want to submit a bug report or patch for a subsystem with they are unfamiliar with.
But the thought of a database of "how to start a good patch title" is far-fetched. Really, as a patch author, just look how other people write patch titles and judge whether this is good for your work too or not.
Either the maintainer wants the patch. Then he is certainly able to apply it no matter the subject line (I personally am getting a lot of patches which don't follow the format I am using in my tree ... converting Subject: lines is so trivial that I have never felt like bothering anyone about it ... it's basically single condition in a shellscript). Or the
It's slightly more than that if you're dealing with more than one area, and I also find this sort of stuff is a good flag for scrubbing the patch in greater detail - when patches stand out from a 1000ft visual overview there's a fair chance that there's other issues so if people regularly submit good patches that have only cosmetic issues I find it's worth guiding them away from that.
On one hand Jiri is right that maintainers can adjust title prefixes ad hoc. (Downside: Weaker connection to mailinglist archives.) On the other hand, in the case of long-term prolific authors like Joe it is more optimal if there is a good patch title right from the outset.
So, if this boring thread does at least yield the conclusion that ${path}/${filename}: is a bad title prefix, at least something was won. :-)
Another thought: Whether a typical part of a mass conversion, e.g. to use a new helper macro without change of functionality, is named
[PATCH] [subsystem] driver: use foo_bar helper or [PATCH] use foo_bar helper in subsystem, driver
does not really matter, does it? This change is more about the helper than about the driver. It is really a different kind of changeset than a functional change that we want to be called
[PATCH] [subsystem] driver: fix crash at disconnection
or so. This is something that those who look for release notes of that driver or subsystem want to grep in the changelog.
Or in other words: If you as patch author wonder what would be a good title for your patch, then ask yourself: How should this change show up in kernel release notes that are constructed from the git shortlog? Sometimes the answer to this question includes among else a prefix with a canonical subsystem name (even case sensitive, with brackets or colon), whereas other times such formalities are utterly pointless.
[Sorry for the spent electrons. But OTOH, issues like (1.) optimum use of reviewer bandwidth, (2.) kernel changelog alias release notes /do/ matter.]