[quagga-dev 7635] Re: [PATCH 05/10] ospfd: remove unneeded memset from a very hot function

paul at jakma.org paul at jakma.org
Thu Jan 14 15:35:24 GMT 2010


On Mon, 11 Jan 2010, Joakim Tjernlund wrote:

> So measure that now when you are at it if you really need to have a 
> profile for every little improvement.

This one seems clear, and someone will integrate it when they have 
time.

More generally, my point was that I'd like to try raise the standard 
for supporting evidence submitted with patches, whenever there could 
be any doubt about the cost/risks/benefits of a patch, e.g.:

- unit tests (separate out stateless logic to make this easier)
- functional testing ("havnt tested this at all!" is common
   enough, but the other extreme is very rare)
- any other useful objective material, like profile where a
   micro-patch claims to improve performance.

The only way development can scale is if these responsibilities are 
spread out.

If instead it's left mostly to the maintainers to determine evidence 
of costs/risks/benefits then the process doesn't scale, and the 
average time a patch gets to sit and wait will go up with the amount 
of patches submitted (presuming the maintainers don't get any extra 
time pressures!). Or the canonical branch gets crap-flooded. ;)

Particularly for bigger patches, this kind of evidence is critical if 
the code is to be maintainable in the long run, and the quality is to 
be maintained.

Oh, I suspect you think we ignore some of your patches, but that's 
not it. If you poke around my git repositories, I've got a fair few 
patches that just havn't been tested enough to get in, even though 
they /probably/ improve matters (e.g. reform of the ospf flooding 
paths, to name one).

NB: This email isn't aimed at you especially ;)

regards,
-- 
Paul Jakma	paul at jakma.org	Key ID: 64A2FF6A
Fortune:
we're waiting for [the phone company] to fix that line



More information about the Quagga-dev mailing list