hgwebdir performance: hgeol

Michał Sznajder michalsznajder at gmail.com
Tue Jan 17 19:36:39 UTC 2012


On Tue, Jan 17, 2012 at 2:24 AM, Antoine Pitrou <solipsis at pitrou.net> wrote:
> Antoine Pitrou <solipsis <at> pitrou.net> writes:
>>
>> Mmmh, thanks for the suggestion. However, judging by the following profile,
>> hard-coding the paths would probably not change much if anything:
>>
>> Tue Jan 17 01:29:08 2012    hgserve.prof
>>
>>          465738 function calls (456430 primitive calls) in 24.582 CPU seconds
>>
>>    Ordered by: internal time
>>    List reduced from 926 to 30 due to restriction <30>
>>
>>    ncalls  tottime  percall  cumtime  percall filename:lineno(function)
>>        35   16.855    0.482   16.855    0.482 {select.select}
>>       163    6.234    0.038    6.234    0.038 {mercurial.parsers.parse_index2}
>
> And it seems the culprit for eagerly parsing the revlogs is hgeol:
>
> [snip]
>  File "/data/hg/lib/python/hgext/eol.py", line 288, in loadeol
>    eol = parseeol(self.ui, self, nodes)
>  File "/data/hg/lib/python/hgext/eol.py", line 208, in parseeol
>    data = repo[node]['.hgeol'].data()
> [snip]

Why do you need hgeol enabled on server? For me EolExntesions seem to
be completely "client" oriented.
Unless you are using eol.checkheadshook hook.

Michal



More information about the Mercurial mailing list