Re: [Monotone-devel] non_workspace_keydir, again

[Stephen Leake]

Stephen Leake <address@hidden> writes:

> I just noticed that the non_workspace_keydir test is failing on my
> MinGW32 development computer, but not on the MinGW32 buildbot.
>
> That's because on my computer I have a key in $APPDATA/monotone/keys,
> while the buildbot does not.
>
> So I set $APPDATA in the test; that fixes it.
>
> However, I noticed there are commented out tests:
>
> -- this should find a private key in the specified keydir
>
> -- ...
>
> -- srv = bg(pure_mtn("serve", "--confdir="..test.root, 
> "--keydir="..test.root.."/keys"), expected_ret, false, true)
> -- sleep(2)
> -- srv:finish()
> -- check(qgrep("beginning service", "stderr"))
>
> -- this should find a private key in the keys directory under the specified 
> confdir
>
> -- srv = bg(pure_mtn("serve", "--confdir="..test.root), expected_ret, false, 
> true)
> -- sleep(2)
> -- srv:finish()
> -- check(qgrep("beginning service", "stderr"))
>
>
> Should these be restored? They pass when I uncomment them on my
> MinGW32 machine.

I uncommented and checked in. These tests cause failures on the
following buildbots:

x86-linux-glibc
x86-linux-glibc-icc
x86-openbsd
i386-suse93

The error is the same on all:

mtn: network error: bind(2) error: Address already in use

This test passes on the other buildbots, so it seems like something
peculiar to those buildbots.

-- 
-- Stephe