[Bug 33858] Multiple 32-bit applications crash due to virtual address space exhaustion (Far Cry 3, Splinter Cell: Blacklist, The Testament of Sherlock Holmes)

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=33858

--- Comment #21 from Zebediah Figura <z.figura12 at gmail.com> ---
(In reply to Paul Gofman from comment #20)
> (In reply to Zebediah Figura from comment #19)
> 
> > I'm not sure I see how it's useful, given what I said above—ways to relieve
> > address space pressure, if they even exist, may be specific to each
> > application affected by exhaustion. 
> 
> I don't see how the ways to relieve address space pressure are specific to
> applications, at least the ones gathered here now. The common problem is
> that Wine needs more virtual address space than Windows, mostly due to
> native libraries often eating more address space (with pulseaudio and opengl
> somewhere on top). Maybe some Wine components could use optimization in that
> regards too.

If it's things like PulseAudio or OpenGL, there's broadly little we can do, and
in that case it may have to be resolved WONTFIX. But sometimes specific Wine
components eat more address space than they need to.

Bottom line, I guess, sometimes excessive address space use is a bug (e.g. a
leak) and can be fixed; sometimes it's a matter of one of many parts of Wine
using more memory than it really needs, sometimes this regarding one or more
system components, sometimes it's all necessary overhead related to API
translation. Usually PulseAudio and OpenGL are by far the biggest culprit,
sure, but not always.

> 
> It is not a single localized issue in Wine, yes, but what specifics do you
> see with regards to applications gathered here? And if is anyone going to do
> something in Wine or some libraries to attempt to relief memory pressure a
> bit, isn't it easier that the information is linked to the single place, so
> it is easier to test with different applications? 

Nothing with these applications, but of course I haven't debugged any of them.
It may be that all of the applications mentioned here can't be helped by
anything in Wine.

I'm personally ambivalent as to whether metabugs are useful in general, though
the policy here seems to be to avoid them. In the case of address space
exhaustion, though, I don't know what there is to test. The problem is quite
simple: too much address space is being used. Using less is pretty much
universally good. Probably some applications will be helped, and for others it
won't be enough.

> 
> > A bug titled something like "multiple
> > applications exhaust address space" can't really be closed, not even as
> > WONTFIX.
> 
> Probably a lot of the multiple individual bugs for this cannot be ever
> closed the same way, as whatever can be done will reduce the probability of
> the specific app crash, but will likely not eliminate it completely. If
> there are convincing enough reasons to think that the issue is about the
> same, why having distinct records?

Well, maybe. If there's a leak triggered by a specific application causing
address space exhaustion, that's a bug and can be fixed. Some other things that
are less clearly bugs can probably be fixed. Anything that can't be fixed in
Wine is probably fair to resolve as WONTFIX. But it really does depend on the
application. The fix for a bug with a title like this one would have to fix
*every* application that suffers from address space exhaustion, which is pretty
well impossible, or we'd have to resolve it as WONTFIX and thereby say that
*no* such application can be fixed, which isn't true either (even if it's rare
that there's anything we can fix).

-- 
Do not reply to this email, post in Bugzilla using the
above URL to reply.
You are receiving this mail because:
You are watching all bug changes.