Re: [Plib-devel] Portability - policy suggestion

[Bert D. <dri...@pl...>]

On Sat, 15 Jan 2005, Alex Perry wrote:

> Steve, maybe we could have a policy that, for every apparently-supported
> target architecture, at least one of the developers with write access
> must be able to build and run code on that architecture.  If that is not
> the case, either none of the relevant contributors meets the project's
> criteria for write access or (more likely) they need to be added ...

Hrrrm, with 20/20 hindsight it is obvious that I screwed this one up. 
Let me share what I did.

I reviewed all patches Martin had pointed me to, applied the good ones 
one by one on my development machine, and reported the ones I would not 
take ownership of to this list. After applying a patch, I do a build in 
the development tree to verify that the problem has gone before I commit 
it (for a build problem, that's verifying that the build proceeds, for a 
crash, it's running the appropriate test code, etc).

If I don't have a perfect explanation for why something fixes it, I even 
go as far as temporarily backing out the change in a scratch build tree 
to check that the problem comes back without the fix.

When I'm done, I go back and do a cvs -q diff over the whole tree. I 
obviously borked this step (probably by not running it over the entire 
tree), as the uncommitted patch was happily staring at me when I diffed 
it this morning.

I usually catch such goofs by building a test straight from anon-cvs, 
but that day anon-cvs was a bit backlogged so I couldn't do it at home. 
I did a test build at $ORKPLACE under FreeBSD 4.10 the next day, and as 
fate would have it, the 4.10 has both 
/usr/include/{sys,machine}/soundcard.h, so the patch isn't required 
there.

Anyway, I apologize for the goof. If you can think of procedural changes 
to prevent reoccurance I'm all ears.

For what it's worth, just applying all patches that are available isn't 
the answer. Patch trackers, like the one on Sourceforge, only work when 
patches that are below par get rejected immediately. I've payed dearly 
for keeping patches because they "contain good ideas" -- they linger and 
realistically, if the submitter doesn't find the time to do them right, 
what is the chance that you as a maintainer will?

Cheers,

 				-- Bert

-- 
Bert Driehuis -- dri...@pl... -- +31-20-3116119
If the only tool you've got is an axe, every problem looks like fun!