Frederic, you're completely right that should leak.
Technically, it's not quite a leak---the amount of memory consumed is constant and never grows until you start to use V8---so it should be safe. But definitely irritating. I'll try to have a look is there a simple way to get rid of it altogether (either by the way you suggest or with some other approach). Anyway I'll send you a patch for review if I fix it, but, sorry, I'll put it on low-priority list unless it's worse that I think about it. And just couple of observations. You could just file a bug with the details you found out, for example, this text makes a perfect bug description imho (I've filed http://code.google.com/p/v8/issues/detail?id=654) And if you can point out exact place where you think Dispose method should be referenced, I'll try to fix it as well. thanks again for your report and yours, anton. On Sat, Mar 20, 2010 at 9:42 PM, fcharp <[email protected]> wrote: > Here is an interesting case. As these things go, it is an unlikely > occurrence, but it just happened that it has been interfering with my > development. > > In summary, I have been bothered by the fact that, in certain cases, > Dispose() will not release the GlobalHandles pool_ memory to the > system. The problem is very easy to duplicate: simply create a > program > with the V8 lib. And make Dispose() the only V8-related statement in > the program. On exit, you end up with an 89K memory leak (size > depending on the platform). There are other leaks, but of lesser > importance than this one. > > > What happens is that pool_ was implicitly allocated by the following > line in global-handles.cc: > > > static GlobalHandles::Pool pool_ > > > Of course, this implicit allocation happens even if V8 is not > formally > initialized. When the Dispose() statement is executed, TearDown() is > called and it proceeds to deallocate resources, including the pool_ > chunk -- but only under normal circumstances. And this is where the > problem occurs. In effect, TearDown() checks if the "has_been_setup > flag" is set, denoting that V8 was initialized. In the present case, > the flag is not set because there was no invocation of V8 prior to > the > invocation of Dispose(). As a result, the GlobalHandles::Release() > method is not called by TearDown() and pool_ is not released. > > > I understand that the conditions under which this happens are odd. > But > in my program, I do not always have to start up the V8 engine, but I > always execute Dispose(), just in case. I guess that will teach > me :-). Anyhow, I know how to avoid the problem now, but it may be a > good idea to make sure that TearDown() disposes of all statically > allocated memory blocks, regardless of the state of the > "has_been_setup" flag. I will modify my local files accordingly. > > > Frederic > > -- > v8-dev mailing list > [email protected] > http://groups.google.com/group/v8-dev > > To unsubscribe from this group, send email to > v8-dev+unsubscribegooglegroups.com or reply to this email with the words > "REMOVE ME" as the subject. > -- v8-dev mailing list [email protected] http://groups.google.com/group/v8-dev To unsubscribe from this group, send email to v8-dev+unsubscribegooglegroups.com or reply to this email with the words "REMOVE ME" as the subject.
