Re: Slow stat() on noacl mounts due to x permission check

[Christian Franke <Christian dot Franke at t-online dot de>]

Corinna Vinschen wrote:
> On Feb  3 19:20, Christian Franke wrote:
> > Cygwin stat() may be much slower on noacl mounts than on acl
> > mounts.This is because on noacl mounts the x-permission bit is
> > guessed by checking for "#!", ":" or "MZ" in the first bytes of the
> > file. AFAIKS this is done for all files except *.exe, *.lnk and
> > *.com.
> >
> > A real world testcase with 20120201 snapshot on a (C++/Java
> > development) tree with ~52000 files in ~12000 dirs:
> >
> > .                            noacl  acl
> > find -size ... (after boot)   518s  51s
> > find -size ... (disk cached)   13s   8s
> > find -name ... (disk cached)    3s   3s (does not need stat())
> >
> > Is it really needed to do the header check for each file?
> >
> > Would it break important use cases if the check is only done for
> > files with typical script file extensions?
> > For example: no extension, .sh, .csh, .pl, .py, ...
> >
> > Another more flexible solution would be a mount option to configure this.
> > (xguess=0: no check; 1: some extensions only; 2: all files)
> We already have the exec/notexec/cygexec mount options, see
> http://cygwin.com/cygwin-ug-net/using.html#mount-table

Somehow I misunderstood the notexec option. Sorry for the noise.

Christian


--
Problem reports:       http://cygwin.com/problems.html
FAQ:                   http://cygwin.com/faq/
Documentation:         http://cygwin.com/docs.html
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple