Ian Clatworthy wrote:
> Martin Pool wrote:
>> On 4 February 2010 07:47, Ian Clatworthy <email address hidden> wrote:
>>> Explorer's sub-command launching is reasonably well isolated in
>>> app_runner.py. Perhaps we need a custom implementation of launching
>>> commands on OS X?
>> So every dialog is a separate process? That would explain it, also bug
>> 516447.
>>
>> I guess you either need to find some os x magic to say "these are
>> actually the same app though they're different processes" or run the
>> dialogs in the same process or tolerate the bugs. jfroy may be able
>> to offer some advice.
>>
>
> Or more ideally, a separate thread in the same process. (One advantage
> of the separate processes now is command isolation wrt errors and
> non-blocking of the UI thread).
>
> Ian C.
>
I'll mention that bzrlib isn't particularly good about being used by
multiple threads concurrently. (Witness the recent hot-fix for issues
with Loggerhead on launchpad.)
I guess if you generally use different objects we are mostly ok (we only
have a couple of global caches which we can try to improve on anyway).
But certainly a single Repository object is not thread safe, for example.
John
=:->
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (Cygwin)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
Ian Clatworthy wrote:
> Martin Pool wrote:
>> On 4 February 2010 07:47, Ian Clatworthy <email address hidden> wrote:
>>> Explorer's sub-command launching is reasonably well isolated in
>>> app_runner.py. Perhaps we need a custom implementation of launching
>>> commands on OS X?
>> So every dialog is a separate process? That would explain it, also bug
>> 516447.
>>
>> I guess you either need to find some os x magic to say "these are
>> actually the same app though they're different processes" or run the
>> dialogs in the same process or tolerate the bugs. jfroy may be able
>> to offer some advice.
>>
>
> Or more ideally, a separate thread in the same process. (One advantage
> of the separate processes now is command isolation wrt errors and
> non-blocking of the UI thread).
>
> Ian C.
>
I'll mention that bzrlib isn't particularly good about being used by
multiple threads concurrently. (Witness the recent hot-fix for issues
with Loggerhead on launchpad.)
I guess if you generally use different objects we are mostly ok (we only
have a couple of global caches which we can try to improve on anyway).
But certainly a single Repository object is not thread safe, for example.
John enigmail. mozdev. org/
=:->
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (Cygwin)
Comment: Using GnuPG with Mozilla - http://
iEYEARECAAYFAkt q6jwACgkQJdeBCY SNAAOcBwCbBDIEd sW1O9hJnPk8lW/ 9a1Nm BteLSkOHg7ffX/ n4i
c2QAn1QiaH4y8qL
=ioXp
-----END PGP SIGNATURE-----