LWN: Comments on "API changes under consideration"

API changes under consideration

wli — Sat, 28 Aug 2004 05:18:31 +0000

I'm rather hard-pressed to sympathize with those not submitting their
changes for mainline inclusion or those unable to port their code
properly.

API changes under consideration

pimlott — Fri, 27 Aug 2004 21:04:38 +0000

We love all your patches equally.

API changes under consideration

jmshh — Fri, 27 Aug 2004 08:43:34 +0000

Yes, but limited to code

in the kernel tree
at the time of checking.

There is a lot of stuff not satisfying either condition.

API changes under consideration

wli — Thu, 26 Aug 2004 22:00:09 +0000

Not at all. One can trivially find all callers using grep(1) and update
them.

API changes under consideration

kleptog — Thu, 26 Aug 2004 12:38:39 +0000

What do you mean that compilers cannot detect the change? If they can't, you change the function name. Then you get an error. You can even build an inline function with the old name to do the conversion.

Not changing the function name just seems like asking for trouble.

API changes under consideration

wli — Thu, 26 Aug 2004 04:48:40 +0000

You've got to be kidding me! I wrote sexier patches than this, and ones
that actually got somewhere! What about the all-new ultra-lightweight
O(1) proc_pid_statm() implementation for 2.6 /proc/ semantics? Or the
/proc/profile consolidation, and the /proc/profile livelock fixes I wrote
for 512x Altixen atop that? Or the CLONE_IDLETASK removal that fixed the
init_idle() cleanups so they booted on sparc32 and sparc64 and optimized
all bootstrap-related code out of runtime fork() codepaths entirely?