win32 fixes (was [BUG] error in HTTPTestUtil.py)

[John A Meinel]

Alexander Belchenko wrote:
> John A Meinel wrote:
>> Well, I went ahead and setup my win32 branch, available from:
>> http://bzr.arbash-meinel.com/branches/bzr/win32/
> 
> It could be rsynced or only bzr branch from this location?

Unfortunately, just bzr branch. But you can copy a recent "bzr.dev" tree
into a new directory, and then pull over it:
  cp -ar bzr.dev bzr-win32
  cd bzr-win32
  bzr pull --remember --overwrite \
      http://bzr.arbash-meinel.com/branches/bzr/win32/

> 
>> As I am logged into windows right now, and wanted to get the tests
>> passing. I know Aaron merged the critical patches (and Martin merged
>> that into mainline for 0.6),
>> so are there specific patches which still need to be applied?
>>
>> I did apply the HTTP one, and my earlier one for APPDATA (plus cleaning
>> that up a bit).
>>
>> It looks like at least your Patch #1 has already been merged.
> 
> Well, I'm update my patches page right now and remove links to already
> included patches. Patch #11 is yours job for HTTPTestUtil.

I've merged the entries that I found were necessary so far. There might
be a couple more, but some of it I'm more interested in fixing from the
inside, so your specific patch may not work.

I believe 1 and 2 were already merged,
3 you removed,
4 was for the older form (and I fixed somewhat with my APPDATA patch).
5 I merged,
6 I did in a different way,
7 I'm waiting on, because I'm thinking it should accept that form for
specific_files.
8 wasn't required yet
9 I merged, though I want to think more about how to properly disable
the executable bit on win32. Otherwise bzr will try and unset the
executable bit.
10 I fixed in a slightly different way
11 I already added
12 I merged
13 I merged, but I'm thinking about removing it. I'm not unhappy with
it, but it isn't necessarily wrong to wait a little bit for another
program to release a lock. And see my other emails about how lock
handling needs to be fixed in bzr.

> 
> Patch #12 is important because it have efffect on plugin subsystem.
> Other patches is for selftest.
> 
> I have one additional patch for bzrlib but I need to update them to
> current bzr.dev state. It partially may affects on config and log
> subsystem.
> 
> I have 2 additional questions.
> 
> 1. Where we should place .bzr.log file on Windows? I mean where should
> be *right* place?

I was trying to think about that myself. I'm thinking that $TMP/bzr.log
might actually be the best place.
Aside from that, I would put it back into $APPDATA/bazaar/2.0/bzr.log

> 2. How we should pass during some tests that very linux-centric? Per
> example I disable some test that check permission bits as unix-mask and
> some more. But I think for that tests could be created windows
> analogique. What you think about?

For executable bits, we can just skip them, like we are doing right now.
I want to fix the underlying code, so that it doesn't try to remove the
executable bits.

> 
> At now I not fix *all* selftests. My latest from 0.6pre point is:
> 
> Ran 473 tests in 80.125s
> 
> FAILED (failures=3, errors=2)
> 
> ~5-7 tests is disabled.

On my branch, using cygwin, all tests pass (by avoiding the deadlock for
pull).

Under win32 native, there are still a couple tests that fail. I'm
working on other stuff right now, but hopefully I've merged the stuff
that you need, so that you at least have a branch to work from.

I'll try and keep my branch up-to-date against bzr.dev, at least
checking on it 1/day. (I have a cronjob which auto-updates my local copy
of bzr.dev 2/day, and emails me if there is a change).

John
=:->

> 
> 
> 
> Alexander
> 
> 


-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 249 bytes
Desc: OpenPGP digital signature
Url : https://lists.ubuntu.com/archives/bazaar/attachments/20051028/ba9218b7/attachment.pgp