You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@subversion.apache.org by be...@qqmail.nl on 2015/08/20 09:28:24 UTC

RE: JavaHL, 1.9: "Bad file descriptor", "Stream doesn't support thiscapability" errors

Hi Thomas,

The Subversion version of the fix is nominated for 1.9.1
[[
+ * r1696222, r1696225
+   Implement polling on named pipes on Windows to resolve svn+ssh:// problems
+   Justification:
+     Without this patch operations like checkout just fail with an error.
+     This is a regression as older Subversion releases ignored the error.
+   Votes:
]]

Apr just returns EBADF on any attempt to poll on a file or pipe on Windows. This workaround on the Subversion side makes us support polling on pipes. (In Subversion < 1.8 we always assumed that there was no data waiting, because we never bothered to check for an error)

The apr version of the patch is harder than I expected, because the poll function supports many flags of which we can really only implement one new variant, that can’t really be combined with any other flags.


We need one more vote on this, before we can start the release process on 1.9.1.

I would hope we can also get r1696695 in the release, as that problem appears to break Subversion for at least some Linux distributions.



We might want to consider also fixing the polling in 1.8 (and 1.7?), as that would allow exiting much earlier when a commit fails. Currently Subversion (when using svn+ssh:// on Windows) just goes on trying to commit on Windows and only notices that svnserve returned an error when done.


Bert

Sent from Mail for Windows 10



From: Thomas Singer
Sent: donderdag 20 augustus 2015 08:23
To: dev@subversion.apache.org
Subject: Re: JavaHL, 1.9: "Bad file descriptor", "Stream doesn't support thiscapability" errors


Hi Bert,

> I'll see if I can properly fix this (preferably in both Subversion and Apr), without reverting to just ignoring errors.

Did you have any success fixing it yet? Is there already a bug entered 
in the issue tracker? I'd rate it as blocker priority.

--
Best regards,
Thomas Singer
=============
syntevo GmbH
http://www.syntevo.com
http://www.syntevo.com/blog



Re: JavaHL, 1.9: "Bad file descriptor", "Stream doesn't support thiscapability" errors

Posted by Thomas Singer <th...@syntevo.com>.
> The fix is alread on the 1.9.x branch and will be
> part of the 1.9.1 release

Perfect, I can confirm that it is working now. Thanks.

-- 
Tom


On 25.08.2015 12:39, Branko Čibej wrote:
> On 25.08.2015 12:36, Thomas Singer wrote:
>> Is this bug already reported in the issue tracker? I've searched but
>> could not found. Should I report it?
>
>
> We don't need that. The fix is alread on the 1.9.x branch and will be
> part of the 1.9.1 release, which we'll probably expedite precisely
> because of this bug.
>
> -- Brane
>
>
>>
>> --
>> Best regards,
>> Thomas Singer
>> =============
>> syntevo GmbH
>> http://www.syntevo.com
>> http://www.syntevo.com/blog
>>
>>
>> On 20.08.2015 09:28, bert@qqmail.nl wrote:
>>> Hi Thomas,
>>>
>>> The Subversion version of the fix is nominated for 1.9.1
>>> [[
>>> + * r1696222, r1696225
>>> +   Implement polling on named pipes on Windows to resolve svn+ssh://
>>> problems
>>> +   Justification:
>>> +     Without this patch operations like checkout just fail with an
>>> error.
>>> +     This is a regression as older Subversion releases ignored the
>>> error.
>>> +   Votes:
>>> ]]
>>>
>>> Apr just returns EBADF on any attempt to poll on a file or pipe on
>>> Windows. This workaround on the Subversion side makes us support
>>> polling on pipes. (In Subversion < 1.8 we always assumed that there
>>> was no data waiting, because we never bothered to check for an error)
>>>
>>> The apr version of the patch is harder than I expected, because the
>>> poll function supports many flags of which we can really only
>>> implement one new variant, that can’t really be combined with any
>>> other flags.
>>>
>>>
>>> We need one more vote on this, before we can start the release
>>> process on 1.9.1.
>>>
>>> I would hope we can also get r1696695 in the release, as that problem
>>> appears to break Subversion for at least some Linux distributions.
>>>
>>>
>>>
>>> We might want to consider also fixing the polling in 1.8 (and 1.7?),
>>> as that would allow exiting much earlier when a commit fails.
>>> Currently Subversion (when using svn+ssh:// on Windows) just goes on
>>> trying to commit on Windows and only notices that svnserve returned
>>> an error when done.
>>>
>>>
>>> Bert
>>>
>>> Sent from Mail for Windows 10
>>>
>>>
>>>
>>> From: Thomas Singer
>>> Sent: donderdag 20 augustus 2015 08:23
>>> To: dev@subversion.apache.org
>>> Subject: Re: JavaHL, 1.9: "Bad file descriptor", "Stream doesn't
>>> support thiscapability" errors
>>>
>>>
>>> Hi Bert,
>>>
>>>> I'll see if I can properly fix this (preferably in both Subversion
>>>> and Apr), without reverting to just ignoring errors.
>>>
>>> Did you have any success fixing it yet? Is there already a bug entered
>>> in the issue tracker? I'd rate it as blocker priority.
>>>
>>> --
>>> Best regards,
>>> Thomas Singer
>>> =============
>>> syntevo GmbH
>>> http://www.syntevo.com
>>> http://www.syntevo.com/blog
>>>
>>>
>>>
>
>

Re: JavaHL, 1.9: "Bad file descriptor", "Stream doesn't support thiscapability" errors

Posted by Branko Čibej <br...@wandisco.com>.
On 25.08.2015 12:36, Thomas Singer wrote:
> Is this bug already reported in the issue tracker? I've searched but
> could not found. Should I report it?


We don't need that. The fix is alread on the 1.9.x branch and will be
part of the 1.9.1 release, which we'll probably expedite precisely
because of this bug.

-- Brane


>
> -- 
> Best regards,
> Thomas Singer
> =============
> syntevo GmbH
> http://www.syntevo.com
> http://www.syntevo.com/blog
>
>
> On 20.08.2015 09:28, bert@qqmail.nl wrote:
>> Hi Thomas,
>>
>> The Subversion version of the fix is nominated for 1.9.1
>> [[
>> + * r1696222, r1696225
>> +   Implement polling on named pipes on Windows to resolve svn+ssh://
>> problems
>> +   Justification:
>> +     Without this patch operations like checkout just fail with an
>> error.
>> +     This is a regression as older Subversion releases ignored the
>> error.
>> +   Votes:
>> ]]
>>
>> Apr just returns EBADF on any attempt to poll on a file or pipe on
>> Windows. This workaround on the Subversion side makes us support
>> polling on pipes. (In Subversion < 1.8 we always assumed that there
>> was no data waiting, because we never bothered to check for an error)
>>
>> The apr version of the patch is harder than I expected, because the
>> poll function supports many flags of which we can really only
>> implement one new variant, that can’t really be combined with any
>> other flags.
>>
>>
>> We need one more vote on this, before we can start the release
>> process on 1.9.1.
>>
>> I would hope we can also get r1696695 in the release, as that problem
>> appears to break Subversion for at least some Linux distributions.
>>
>>
>>
>> We might want to consider also fixing the polling in 1.8 (and 1.7?),
>> as that would allow exiting much earlier when a commit fails.
>> Currently Subversion (when using svn+ssh:// on Windows) just goes on
>> trying to commit on Windows and only notices that svnserve returned
>> an error when done.
>>
>>
>> Bert
>>
>> Sent from Mail for Windows 10
>>
>>
>>
>> From: Thomas Singer
>> Sent: donderdag 20 augustus 2015 08:23
>> To: dev@subversion.apache.org
>> Subject: Re: JavaHL, 1.9: "Bad file descriptor", "Stream doesn't
>> support thiscapability" errors
>>
>>
>> Hi Bert,
>>
>>> I'll see if I can properly fix this (preferably in both Subversion
>>> and Apr), without reverting to just ignoring errors.
>>
>> Did you have any success fixing it yet? Is there already a bug entered
>> in the issue tracker? I'd rate it as blocker priority.
>>
>> -- 
>> Best regards,
>> Thomas Singer
>> =============
>> syntevo GmbH
>> http://www.syntevo.com
>> http://www.syntevo.com/blog
>>
>>
>>


Re: JavaHL, 1.9: "Bad file descriptor", "Stream doesn't support thiscapability" errors

Posted by Thomas Singer <th...@syntevo.com>.
Is this bug already reported in the issue tracker? I've searched but 
could not found. Should I report it?

--
Best regards,
Thomas Singer
=============
syntevo GmbH
http://www.syntevo.com
http://www.syntevo.com/blog


On 20.08.2015 09:28, bert@qqmail.nl wrote:
> Hi Thomas,
>
> The Subversion version of the fix is nominated for 1.9.1
> [[
> + * r1696222, r1696225
> +   Implement polling on named pipes on Windows to resolve svn+ssh:// problems
> +   Justification:
> +     Without this patch operations like checkout just fail with an error.
> +     This is a regression as older Subversion releases ignored the error.
> +   Votes:
> ]]
>
> Apr just returns EBADF on any attempt to poll on a file or pipe on Windows. This workaround on the Subversion side makes us support polling on pipes. (In Subversion < 1.8 we always assumed that there was no data waiting, because we never bothered to check for an error)
>
> The apr version of the patch is harder than I expected, because the poll function supports many flags of which we can really only implement one new variant, that can’t really be combined with any other flags.
>
>
> We need one more vote on this, before we can start the release process on 1.9.1.
>
> I would hope we can also get r1696695 in the release, as that problem appears to break Subversion for at least some Linux distributions.
>
>
>
> We might want to consider also fixing the polling in 1.8 (and 1.7?), as that would allow exiting much earlier when a commit fails. Currently Subversion (when using svn+ssh:// on Windows) just goes on trying to commit on Windows and only notices that svnserve returned an error when done.
>
>
> Bert
>
> Sent from Mail for Windows 10
>
>
>
> From: Thomas Singer
> Sent: donderdag 20 augustus 2015 08:23
> To: dev@subversion.apache.org
> Subject: Re: JavaHL, 1.9: "Bad file descriptor", "Stream doesn't support thiscapability" errors
>
>
> Hi Bert,
>
>> I'll see if I can properly fix this (preferably in both Subversion and Apr), without reverting to just ignoring errors.
>
> Did you have any success fixing it yet? Is there already a bug entered
> in the issue tracker? I'd rate it as blocker priority.
>
> --
> Best regards,
> Thomas Singer
> =============
> syntevo GmbH
> http://www.syntevo.com
> http://www.syntevo.com/blog
>
>
>