[darcs-users] darcs patch: Update documentation to reflect GHC 6.6 ... (and 1 more)

Tommy Pettersson ptp at lysator.liu.se
Fri Jan 9 21:11:45 UTC 2009


On Sun, Jan 04, 2009 at 03:25:53PM +1100, Trent W. Buck wrote:
> The caveat was originally added by the following changeset; Tommy, are
> you still around and do you have an opinion on what, if anything, should
> go here?

I'm around, but not very often, and I no longer read all emails
to darcs-users.

I no longer remember any details, but this is from an old email
I found (written by me):

| The problem is with the C locale.  Then Haskell's isPrint
| should works just fine, and it does ... except for three users
| having reported problems, all when upgrading to GHC 6.4.  But
| when I recently upgraded to GHC 6.4 it worked, of course... ;-/
| So there must be something else in combination with GHC 6.4.
| 
| Mark Stosberg had the problem on FreeBSD 4.10 with perl 5.8.0.
| I didn't get it on Linux 2.6 with perl 5.8.7.  I can't find
| the other mails now.
| 
| The precise problem is (as far as I understands it): perl sets
| LC_ALL to C, perl invokes darcs, in darcs isPrint erroneously
| returns true for chars in the range \x80 (or maybe \xa0)
| to \xff.
| 
| It could be GHC, perl or even libc that wrongs.

This was three years ago, so my memory is faint, but it's
starting to come back. I think the above message was about a
patch I made with a foreign call to c_isprint, because we had a
failing test. But it didn't help, and so it was not applied.

The test must have been removed, but the test printer.sh checks
that all the funny escapes work *without* any of the
configurable exceptions, so if we set DARCS_USE_ISPRINT in that
test, and run it on all the buildbots, we could check if any of
these systems still triggers the bug.


-- 
Tommy Pettersson <ptp at lysator.liu.se>


More information about the darcs-users mailing list