[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Terrific Dead Lock

From: Neil Jerram
Subject: Re: Terrific Dead Lock
Date: Mon, 17 Mar 2008 23:20:29 +0000
User-agent: Gnus/5.110006 (No Gnus v0.6) Emacs/21.4 (gnu/linux)

address@hidden (Ludovic Courtès) writes:

> Hi,
> address@hidden (Ludovic Courtès) writes:
>> Two possibilities to fix it:
>>   1. Copy `srcprop.[ch]' and `eval.c' bits from HEAD to 1.8.  After all,
>>      it's probably solid enough (I use almost only HEAD).  See [0] for
>>      an overview of the initial patch.  It doesn't break the public API
>>      nor the ABI, but it (re)moves stuff from the `srcprop.h'.
>>   2. Remove the critical section from 1.8 and synchronize accesses to
>>      `srcprops_freelist' with a private mutex, assuming that's a correct
>>      fix.

I prefer (1).

> The second approach is actually wrong: we'd need to acquire a mutex in
> `srcprops_free ()' to synchronize accesses to `srcprops_freelist', but
> we can't do it since `srcprops_free ()' is called during GC---which is
> why there was a critical section in the first place I suppose.
> Thus the first approach seems unavoidable.  Attached is the exact patch.
> It removes non-public macros and data types from `srcprop.h', which is
> acceptable IMO.
> OK to apply?

Basically yes, but two thoughts:

- Can I just take a couple more days to review the srcprops changes in
  detail?  It's important for my debugging work, and I recall having
  some concern when Han-Wen implemented this...  Looking at the diffs
  quickly again now, I can't see any reason for that concern, but I'd
  like to be sure.

- Although I agree in principle that all those macros shouldn't be in
  srcprop.h, do we really need to make that change now?  I think we can
  just apply Han-Wen's changes to the macros in srcprop.h, without
  moving them to srcprop.c.


reply via email to

[Prev in Thread] Current Thread [Next in Thread]