Georg Baum <[EMAIL PROTECTED]> writes:
| Lars Gullik Bjønnes wrote:
>
>> Georg Baum <[EMAIL PROTECTED]> writes:
>> | - Exit lyx from the buffer constructor if the temp dir is invalid. I
>> | tested it (using a static counter that triggered the test in the third
>> | buffer), and unsaved changes got written to .emergency files.
>> | I guess that doing this "controlled crash" is safe: The new buffer is
>> | completely constructed, but empty, so the emergency write will not harm.
>>
>> I don't agree with this one. Sure we should do a _controlled_ exit if
>> we are unable to create the tempdir, And we should notify the user if
>> we are unable to create the buffertmpdir, but we should not just exit.
>
| The user is notified.
And replace the but with an "and"
>
>> If the only way to do this properly is to use exceptions then let's do
>> just that. And have the no-exception case exit with a lousy user
>> message.
>
| Of course I am not the one to decide this. And I don't want to put too much
| work into this, since the primary goal is to make the "include material"
| insets working in all cases, and removing the no temp dir case is only
| something that makes this easier.
| May I suggest to just ignore the case that createBufferTmpDir() could fail
| for now and only put in a FIXME?
I can live with that. But I cannot live with the exit.
| Then I wonder why there is a per buffer temp dir at all? This problem would
| vanish if we only had the global temp dir.
then the deps file must be renamed.
(and perhaps other files..., probably most.)
| The doxygen output (version 1.3.4) looks ok.
So I am the one confused...
--
Lgb
