[lustre-devel] [RFC PATCH 0/3] staging: lustre: detypedef

Joe Perches joe at perches.com
Fri Apr 1 12:27:23 PDT 2016


On Fri, 2016-04-01 at 19:14 +0000, Drokin, Oleg wrote:
> On Apr 1, 2016, at 2:44 PM, Joe Perches wrote:
> > On Fri, 2016-04-01 at 14:23 +0000, Drokin, Oleg wrote:
> > > On Apr 1, 2016, at 9:02 AM, Joe Perches wrote:
> > > > Question about removing lustre typedefs.
> > > > 
> > > > Various bits of lustre code use a mix of struct foo and foo_t.
> > > > 
> > > > When would be an appropriate time to submit patches similar to
> > > > below that individually remove various typedefs from lustre code?
> > > I think now is as good time as any.
> > > the only small correction is those are LNet typedefs.
> > > While LNet is technically part of Lustre, it's a bit of a separate
> > > thing useful without Lustre too.
> > > 
> > > I know James is working on cleaning up LNet, but I don't know if he has
> > > anything this would be conflicting at this moment or not.
> > > 
> > > Thanks for the patches. I wonder if you are generating them automatically?
> > > Because it would be great if it also fixes the alignment issues
> > It's pretty automatic.
> > 
> > It's a trivial variant of the detypedef perl script I wrote awhile ago:
> > http://permalink.gmane.org/gmane.linux.drivers.driver-project.devel/18603
> > 
> > I think changing the alignment issues is better done in a
> > separate patch.
> but then it's two patches per change in a way. fixing one thing breaking
> the other warning-wise, that's why I typically try to make such cleanup
> patches not to introduce any new warnings.

detypedef frequently introduces > 80 column lines.

It's much easier to verify this way without introducing changes
like rewrapping to 80 column.




More information about the lustre-devel mailing list