[PATCH 02/17] Let calloc handle multiplication

walter harms wharms at bfs.de
Thu Apr 2 07:03:16 PDT 2015

Am 01.04.2015 20:36, schrieb Alan Coopersmith:
> On 04/ 1/15 02:53 AM, walter harms wrote:
>> Why not generally call xnfmalloc/calloc/realloc and exit on OOM ?
>> The strategy now seems otherwise.
> Because this is the X server and taking down your entire session because
> one client made a bad request is very very user unfriendly and will result
> in all users screaming their heads off and demanding we revert the change
> or they will all go off and use something else and boy we'll be sorry then.
> We generally use the xnf* interfaces in places like server setup which
> aren't
> handling client requests and where we can't just continue on handling other
> clients and other requests if we fail.

ok i see, that was not obvious from the diff

>> No problem we me, in the real world i never encountered programs that
>> try to handle situations like OOM except doing exit(1).
> Clearly you've never looked at sources for robust & reliable software like
> databases that can't just say "Oh crap!  Got an error!  Destroying data
> now!"
To be fair, the last time i managed a database into OOM, i wished it had. Instead
it crashed and caused a lot work ...
NTL normally i do client stuff


More information about the xorg-devel mailing list