Re: [PATCH] history locking with fcntl

[Vincent Lefevre <vincent@xxxxxxxxxx>]

On 2008-04-17 09:23:07 -0700, Wayne Davison wrote:
> I fixed the bogus "bad file descriptor" part of that error back on March
> 6th, as well as making some minor improvements to the errors.  I'd like
> to see what the real error is when your flock patch is not present.

Note: it's fcntl, not flock (flock does not lock over NFS).

If I do not set the option (which is equivalent to not using the patch),
then I get immediate history corruption when using two machines (well,
at least 64-bit ones), 100% reproducible.

> Attached is an alternative patch for the Src/utils.c file that
> implements this (it is easier to see what I've done than diffing it
> against the file with the current fcntl changes).  Let me know what
> you think.

I'll try it.

> Additional discussion:  I think the option name should be made more
> generic so that it can be made to support more types of file locking,
> such as the flock() call.  Perhaps call it HIST_EXTRA_LOCKING?

Yes, perhaps. But this would give the user less control on the lock
mechanisms.

-- 
Vincent Lefèvre <vincent@xxxxxxxxxx> - Web: <http://www.vinc17.org/>
100% accessible validated (X)HTML - Blog: <http://www.vinc17.org/blog/>
Work: CR INRIA - computer arithmetic / Arenaire project (LIP, ENS-Lyon)