>>>>> John A Meinel <email address hidden> writes:
<snip/>
> Should we be forcing "utf-8" rather than _fs_enc?
There are other similar bugs where we trigger unicode errors because the
fs encoding ends up being ascii (generally because the user didn't
properly specify the encoding in the env) and the end result is that we
more or less agreed to go directly to utf8.
> I would tend to do
> 1) _fs_enc if set (not None) and not ASCII
But in which cases will that be different than utf8 (especially for a
home directory) ? And is it worth the trouble given that we never got
bug reports about not encoding BZR_HOME ?
> 2) fallback to 'mbcs' on Windows
mbcs is used already for decoding BZR_HOME on windows (even with this
patch).
> 3) fallback to 'utf-8' everywhere else
Given that this test is failing on *all* slaves (not only the Ubuntu
ones), I'd rather land this than bikeshed too long for unclear benefits.
>>>>> John A Meinel <email address hidden> writes:
<snip/>
> Should we be forcing "utf-8" rather than _fs_enc?
There are other similar bugs where we trigger unicode errors because the
fs encoding ends up being ascii (generally because the user didn't
properly specify the encoding in the env) and the end result is that we
more or less agreed to go directly to utf8.
> I would tend to do
> 1) _fs_enc if set (not None) and not ASCII
But in which cases will that be different than utf8 (especially for a
home directory) ? And is it worth the trouble given that we never got
bug reports about not encoding BZR_HOME ?
> 2) fallback to 'mbcs' on Windows
mbcs is used already for decoding BZR_HOME on windows (even with this
patch).
> 3) fallback to 'utf-8' everywhere else
Given that this test is failing on *all* slaves (not only the Ubuntu
ones), I'd rather land this than bikeshed too long for unclear benefits.