MQ performance on large repo

Benoit Boissinot bboissin at gmail.com
Wed Mar 3 12:33:59 CST 2010


On Wed, Mar 3, 2010 at 7:25 PM, Greg Ward <greg at gerg.ca> wrote:
> Oh yeah: nobody has asked one question that has been bugging me: *why*
> does strip need to read fncache?  I mean, it's nice that this change
> means reading fncache is now roughly 10x faster (because the file is
> 10x smaller), but it still seems odd.
>
> I've followed the stack trace down and every step along the way looks
> rational on its own ... but when you step back and look at the big
> picture, you just have to ask yourself: why does strip need to know
> the list of decoded filelog filenames?

store is pretty simple right now, it needs to add new filename so it
catches every opener() calls. It could probably be more clever and do
it only when the file is new instead (file does not exist, or size is
zero).

regards,

Benoit


More information about the Mercurial-devel mailing list