I enabled this in "always compile with mmap support", but it seems
actually using mmap in Darcs.Util.ByteString causes test failures
on Windows, e.g. in issue1248. I'm not sure if I didn't run the test
suite at the time of my original patch or if something has changed
on my machine, but either way it seems dubious to keep using mmap there
until we know when it's safe to do.

In this patch I've directly inlined the "no mmap" behaviour choice
and made it Windows only rather than passing a flag through from
cabal as was done previously, as I'm not aware of non-Windows cases
where it's needed.

Note that we still depend on the mmap package for hashed-storage
which has always required it. I'm not sure why or if it's ok for
hashed-storage to be using it.


1 patch for repository darcs-unstable@darcs.net:screened:

Sat Sep 19 10:26:40 BST 2015  Ganesh Sittampalam <ganesh@earth.li>
  * using mmap on Windows was causing test failures

Attachments

• patch-preview.txt
• unnamed
• using-mmap-on-windows-was-causing-test-failures.dpatch

self-accepting as it's been waiting for review for a few weeks - also 
it's pretty trivial

I'm pushing this to branch 2.10 so more people can have working tests.