[GEM-dev] "couldn't create" error and can't save patches

[IOhannes m zmoelnig]

On 2010-03-11 05:32, Stephen Lucas wrote:
> 
> When this occurs, I am unable to load any more abstractions (dynamically
> or manually) but I am able to create regular objects. However, it is always
> accompanied by an inability to save, which produces the error:
> 
> error: (save path): Bad file descriptor

sounds to me like the famous file-handle leak on w32.
could you try to run the task manager, enable the display of "number of
handles" column in the "processes" tab, and watch it rise whenever you
create an object.

> 
> The only way to fix this is to restart pd and hope that I don't lose much
> unsaved work.
> 
> I'm fairly certain that this is a GEM related issue, since I have used the
> same technique to instantiate thousands of non-GEM abstractions without this
> problem. I have not tested to see if the problem happens by creating 70+
> instances manually; however, it is definitely cumulative and the problem
> eventually happens despite clearing and remaking only a few instances.
> 

i have no real idea where this happens; the file-handle leak is known to
occur on w32 when opening e.g. images; is your dynamically-instantiated
abstraction doing something like this?

all of the above should be true for both dynamic and manual patching.

anyhow, it would still be nice to have a minimal patch exposing the
problem (with minimal dependencies as well: only Pd-vanilla and Gem)


the problem is really the way how windows handles file handles: a file
opened in one dll (in our case: pd), can eventually be _not_ closed in
another dll (in our case: Gem). the only solution that i know of, would
involve Pd offering a possibility to also close the handle.


mfgasdr
IOhannes

-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/pkcs7-signature
Size: 3636 bytes
Desc: S/MIME Cryptographic Signature
URL: <http://lists.puredata.info/pipermail/gem-dev/attachments/20100311/a7c55dd2/attachment.bin>