performance problem with hg 1.3 over ssh?

Stuart W. Marks smarks at
Wed Jul 8 19:56:14 CDT 2009

Bryan O'Sullivan wrote:
> On Wed, Jul 8, 2009 at 4:55 PM, Bryan O'Sullivan <bos at 
> <mailto:bos at>> wrote:
>     Per bisect on my Mac, followed by a manual verification, the first
>     bad revision is:
>     changeset:   8340:fce065538bcf
> There's now a fix for this in crew-stable.

Ah, much better, thank you. Performance is now back to what it should be, 
perhaps even 10-15% faster than 1.2.

I had observed my unpatched hg-1.3 doing incredible numbers of 1-byte read() 
calls. This is consistent with the change in fce065538bcf, which left bufsize 
at the default, resulting in (according to pydoc subprocess) unbuffered I/O. 
I'm curious as to why this didn't occur on Linux though.


More information about the Mercurial-devel mailing list