[Unison-hackers] Help building OSX GUI

[Jerome Vouillon]

On Wed, May 06, 2009 at 10:35:30PM -0400, Benjamin Pierce wrote:
[...]
> With -unicode, it works for small sets of files but the update  
> detection on the linux side seems very slow.

I could not reproduce this.

I tried to run Unison on about 100000 files (part of my mail
archives).  This involves sorting large numbers of files, as well as
doing case-insensitive regular expression matching (on the other hand,
the file names are all ASCII, so the normalization function does not
have that much to do).  I used the "immutable" directive, so Unison
does not do much else.  The result is quite good.  The Unicode
case-insensitive mode seems to be actually faster than the old mode,
though this may be just noise.

   Unicode mode
   ./unison -ui text test  1,38s user 0,32s system 99% cpu 1,693 total

   Case insensitive mode
   ./unison -ui text test  1,58s user 0,33s system 100% cpu 1,901 total

   Case sensitive mode
   ./unison -ui text test  0,92s user 0,27s system 81% cpu 1,457 total

   2.32, case sensitive
   ./unison -ui text test  0,91s user 0,29s system 84% cpu 1,432 total

> One funny bug I noticed is that, with -unicode, my "ignore" directives  
> don't work any more!  The pathnames in the ignore directives are  
> clearly not getting normalized the same way as the filenames that it's  
> getting from the OS.

Indeed, this is broken...

-- Jerome