| deadlock at startup in the class EclipseAppContainer [message #488085] |
Fri, 25 September 2009 14:19  |
Eclipse User |
|
|
|
Originally posted by: c.dore.castsoftware.com
Hi
using RCP 3.4.2, on Windows, on a command line (CLI) application.
We are facing a deadlock situation, in a race condition that happens
averagely every 10 times.
the buggy code is in org.eclipse.equinox.internal.app.EclipseAppContainer.
In this class, some methods are synchronized (thus setting a lock on the
EclipseAppContainer instance).
Some other are not synchronized but have parts of it synchronized on the
"apps" instance variable.
Then, these instances are used in several threads, and with that
configuration of locks, some threads lock the EclipseAppContainer
instance then try to lock "apps", some others locks "apps" first then
try to lock the EclipseAppContainer instance.
Thus dead lock, that block our CLI, thus the batches that run it among
other calls.
Unfortunately, we need to deliver quite quickly, so I need to know if
there is a bug filled for that (or do I need to do it, and where) and
most of all if there is any kind of solution for that : workaround,
patch, ...
Thanks for your help.
--
cd
|
|
|
| Re: deadlock at startup in the class EclipseAppContainer [message #488139 is a reply to message #488085] |
Fri, 25 September 2009 16:16  |
| Eclipse User |
|
|
|
Originally posted by: c.dore.castsoftware.com
I just found the bugzilla database for eclipse and filled
https://bugs.eclipse.org/bugs/show_bug.cgi?id=290550
However, any hint/workaround/patch/whatever solution is anyway appreciated
Christophe Dore a écrit :
> Hi
>
> using RCP 3.4.2, on Windows, on a command line (CLI) application.
>
> We are facing a deadlock situation, in a race condition that happens
> averagely every 10 times.
>
> the buggy code is in org.eclipse.equinox.internal.app.EclipseAppContainer.
>
> In this class, some methods are synchronized (thus setting a lock on the
> EclipseAppContainer instance).
>
> Some other are not synchronized but have parts of it synchronized on the
> "apps" instance variable.
>
> Then, these instances are used in several threads, and with that
> configuration of locks, some threads lock the EclipseAppContainer
> instance then try to lock "apps", some others locks "apps" first then
> try to lock the EclipseAppContainer instance.
>
> Thus dead lock, that block our CLI, thus the batches that run it among
> other calls.
>
> Unfortunately, we need to deliver quite quickly, so I need to know if
> there is a bug filled for that (or do I need to do it, and where) and
> most of all if there is any kind of solution for that : workaround,
> patch, ...
>
> Thanks for your help.
>
> --
> cd
|
|
|
Powered by
FUDForum. Page generated in 0.03078 seconds
] 
Search
Help
Register
Login
Home
