slow clone form a windows share

Adrian Buehlmann adrian at cadifra.com
Fri Apr 6 12:02:41 CDT 2012


On 2012-04-05 14:04, marco gaddoni wrote:
> 
> 
> On Thu, Apr 5, 2012 at 09:28, Adrian Buehlmann <adrian at cadifra.com
> <mailto:adrian at cadifra.com>> wrote:
> 
>     On 2012-04-05 08:36, marco gaddoni wrote:
>     > Hello,
> 
>     Googling for "C:\Windows\CSC\v2.0.6" hints that it might be related to
>     Windows' offline files feature.
> 
>     Check your offline files setting and try disabling it:
> 
>     http://www.sevenforums.com/tutorials/48829-offline-files-enable-disable-use.html
> 
> 
> thank you, Adrian,
> offline folders was enabled and following
> that page i was able to disable it;
> 
> unfortunately the speed did not improve.
> this repository contain 2 commit,
> boost 1.48 and 1.49.
> its size is
> 
> server:/share/hg/boost/.hg # du -s *
> 4       00changelog.i
> 8       cache
> 4       hgrc
> 4       requires
> 204756  store
> 0       undo.bookmarks
> 4       undo.branch
> 4       undo.desc
> 0       undo.dirstate
> 
> server:/share/hg/boost/.hg # find store/ | wc -l
> 41070
> 
> server:/share/hg/boost/.hg # find . -type f | xargs du -s | sort -r -n
> 2056    ./fncache
> 1900    ./data/libs/polygon/doc/_g_t_l__boostcon__draft03.pdf.d
> 1184    ./00manifest.d
> 732     ./data/libs/msm/doc/_p_d_f/msm.pdf.d
> 704     ./data/libs/mpl/doc/refmanual.pdf.d
> 440     ./data/libs/pool/doc/pool.pdf.d
> 412     ./data/libs/icl/doc/boostcon09/intro__to__itl__3__0__0__bc09.pdf.d
> 364     ./data/libs/icl/doc/boostcon09/intro__to__itl__3__1__0.pdf.d
> 364     ./data/libs/icl/doc/boostcon09/intro__to__itl.pdf.d
> 328    
> ./data/libs/math/dot__net__example/_setupdistex/_release/_setupdistex.ms
> <http://setupdistex.ms>
> i.d
> 312     ./data/status/expected__results.xml.d
> 244     ./data/more/_boost_sponsorship_agreement.pdf.d
> 244     ./data/libs/math/test/ibeta__inva__data.ipp.d
> 236     ./data/libs/math/test/binomial__quantile.ipp.d
> 224     ./data/libs/statechart/doc/reference.pdf.d
> 224     ./data/libs/python/build/python__v1.zip.d
> 216     ./data/libs/asio/doc/reference.qbk.d
> 208     ./data/libs/geometry/example/data/roads.wkt.d
> 208     ./00changelog.d
> 
> is it normal to take more than 10 minutes to clone ?

How long does it take to (deep-)copy the repo root directory with all
the contents from the network share to your destination? (that is,
without using mercurial).

Copying lots of files from a Windows share to a local computer usually
takes time as well. I wouldn't be that suprised about 10 minutes for a
plain directory copy over the network if the number of files is large
enough.

You most likely can't expect mercurial being faster than a file per file
copy, if it has to access the files an a network share and copy them to
your local computer.

In response to Matt's answer: remember that a local 'hg clone' will
normally be fast because of hardlinking (provided the filesystem
supports it, which NTFS does, and both source and target location are on
the same volume), which isn't possible if you clone from UNC path to
local. So you would probably have to use 'hg clone --pull' to compare
speeds when doing the local clone for comparison.

> i have tried to clone via web (that is, i run hg serve on
> the server and cloned on my pc with
> hg clone http://server:8000/);
> it died in out of memory ...
> 
> is this expected ?

See Matt's answer. I'll refrain from responding to this part.



More information about the Mercurial-devel mailing list