bug#16928: 24.3.50; file _flymake not deleted on remote hosts

bug#16928: 24.3.50; file _flymake not deleted on remote hosts

[Leo Liu]

1. emacs -q
2. (add-hook 'find-file-hook 'flymake-find-file-hook)
3. make sure the php executable is available
4. save https://github.com/drupal/drupal/raw/7.x/update.php in a remote
   host
5. open it in emacs

And I see update_flymake.php not being deleted (cleaned up). So it seems
at the time flymake-simple-cleanup was called, (file-exists-p
"/host:/path/to/update_flymake.php") returned false.

This is kinda annoying because after a while I see lots of _flymake*
files in the project.

Leo

bug#16928: 24.3.50; file _flymake not deleted on remote hosts

[Michael Albinus]

Leo Liu <sdl.web@gmail.com> writes:

> 1. emacs -q
> 2. (add-hook 'find-file-hook 'flymake-find-file-hook)
> 3. make sure the php executable is available
> 4. save https://github.com/drupal/drupal/raw/7.x/update.php in a remote
>    host
> 5. open it in emacs
>
> And I see update_flymake.php not being deleted (cleaned up). So it seems
> at the time flymake-simple-cleanup was called, (file-exists-p
> "/host:/path/to/update_flymake.php") returned false.

Yes, I could reproduce it. The problem is, that flymake starts an
asynchronous process. It's sentinel calls then flymake-simple-cleanup,
which fires file-exists-p to an unpredictable time.

After serving flymake-find-file-hook, find-file has continued with the
other functions in find-file-hook. When the file-exists-p call from the
sentinel happened, Tramp was busy with requests from
vc-find-file-hook. And since Tramp is designed to work sequentially, it
couldn't serve promptly :-(

As temporary workaround, you could try

(add-hook 'find-file-hook 'flymake-find-file-hook 'append)

I will check, whether Tramp could be enabled to handle such asynchronous
calls. 

> Leo

Best regards, Michael.

bug#16928: 24.3.50; file _flymake not deleted on remote hosts

[Leo Liu]

On 2014-03-04 21:14 +0800, Michael Albinus wrote:
> Yes, I could reproduce it. The problem is, that flymake starts an
> asynchronous process. It's sentinel calls then flymake-simple-cleanup,
> which fires file-exists-p to an unpredictable time.
>
> After serving flymake-find-file-hook, find-file has continued with the
> other functions in find-file-hook. When the file-exists-p call from the
> sentinel happened, Tramp was busy with requests from
> vc-find-file-hook. And since Tramp is designed to work sequentially, it
> couldn't serve promptly :-(

I see. Maybe it is better for flymake to just call delete-file without
the file-exists-p check ignoring any error. I have found other cases
where _flymake files are left behind.

> As temporary workaround, you could try
>
> (add-hook 'find-file-hook 'flymake-find-file-hook 'append)

Thanks, will use it for now.

> I will check, whether Tramp could be enabled to handle such
>asynchronous
> calls. 

Thanks.

Leo

bug#16928: 24.3.50; file _flymake not deleted on remote hosts

[Michael Albinus]

Leo Liu <sdl.web@gmail.com> writes:

> I see. Maybe it is better for flymake to just call delete-file without
> the file-exists-p check ignoring any error. I have found other cases
> where _flymake files are left behind.

You could try. But i'm not 100% sure that we don't run into a similar
situation.

However, Tramp's `delete-file' handler doesn't get a special handling in
`vc-find-file-hook', so it might work.

> Thanks.
>
> Leo

Best regards, Michael.

bug#16928: 24.3.50; file _flymake not deleted on remote hosts

[Michael Albinus]

Michael Albinus <michael.albinus@gmx.de> writes:

> Leo Liu <sdl.web@gmail.com> writes:
>
>> I see. Maybe it is better for flymake to just call delete-file without
>> the file-exists-p check ignoring any error. I have found other cases
>> where _flymake files are left behind.
>
> You could try. But i'm not 100% sure that we don't run into a similar
> situation.

I've checked further, there are still problems. A fix seems to be more
complex than expected, it must wait until the trunk is reopened (I fear
I could damage Tramp so close to the release).

For the time being, I have added `tramp-test31-asynchronous-requests' to
Tramp's test cases. It will fail until the bug is fixed.

Best regards, Michael.

bug#16928: 24.3.50; file _flymake not deleted on remote hosts

[Leo Liu]

On 2014-03-07 22:19 +0800, Michael Albinus wrote:
> I've checked further, there are still problems. A fix seems to be more
> complex than expected, it must wait until the trunk is reopened (I fear
> I could damage Tramp so close to the release).
>
> For the time being, I have added `tramp-test31-asynchronous-requests' to
> Tramp's test cases. It will fail until the bug is fixed.

Thank you. It seems best for now.

Leo

bug#16928: 24.3.50; file _flymake not deleted on remote hosts

[Stefan Kangas]

Michael Albinus <michael.albinus@gmx.de> writes:

>>> I see. Maybe it is better for flymake to just call delete-file without
>>> the file-exists-p check ignoring any error. I have found other cases
>>> where _flymake files are left behind.
>>
>> You could try. But i'm not 100% sure that we don't run into a similar
>> situation.
>
> I've checked further, there are still problems. A fix seems to be more
> complex than expected, it must wait until the trunk is reopened (I fear
> I could damage Tramp so close to the release).
>
> For the time being, I have added `tramp-test31-asynchronous-requests' to
> Tramp's test cases. It will fail until the bug is fixed.

Hi Michael,

I'm now seeing the following test pass:

   passed  1/1  tramp-test43-asynchronous-requests (1.738177 sec)

Does that mean that this bug has been fixed?

Best regards,
Stefan Kangas

bug#16928: 24.3.50; file _flymake not deleted on remote hosts

[Michael Albinus]

Stefan Kangas <stefan@marxist.se> writes:

> Hi Michael,

Hi Stefan,

> I'm now seeing the following test pass:
>
>    passed  1/1  tramp-test43-asynchronous-requests (1.738177 sec)

A lot of work has been done in order to let this test pass. It does
almost, but it is still unstable. That's why it is tagged :unstable.

> Does that mean that this bug has been fixed?

I've played the scenario the OP has given several times with Emacs
27.0.50, and the problem didn't happen. So yes, let's assume the bug has
been fixed.

> Best regards,
> Stefan Kangas

Best regards, Michael.

bug#16928: 24.3.50; file _flymake not deleted on remote hosts

[Stefan Kangas]

Michael Albinus <michael.albinus@gmx.de> writes:

> > Does that mean that this bug has been fixed?
>
> I've played the scenario the OP has given several times with Emacs
> 27.0.50, and the problem didn't happen. So yes, let's assume the bug has
> been fixed.

Thanks, I'm closing this now then.  We can always reopen later if this
turns out to still be a problem.

Best regards,
Stefan Kangas