bug-gnu-emacs
[Top][All Lists]
Advanced

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

bug#59426: 29.0.50; [tree-sitter] Some functions exceed maximum recursio


From: Eli Zaretskii
Subject: bug#59426: 29.0.50; [tree-sitter] Some functions exceed maximum recursion limit
Date: Mon, 21 Nov 2022 19:16:06 +0200

> From: Yuan Fu <casouri@gmail.com>
> Date: Mon, 21 Nov 2022 08:52:53 -0800
> Cc: 59426@debbugs.gnu.org
> 
> 
> 
> > On Nov 21, 2022, at 5:19 AM, Eli Zaretskii <eliz@gnu.org> wrote:
> > 
> >> From: Yuan Fu <casouri@gmail.com>
> >> Date: Sun, 20 Nov 2022 16:53:45 -0800
> >> 
> >> 
> >> Emacs crashed on a very large C file when c-ts-mode is on, because
> >> the function building the imenu list tries to walk through the whole
> >> parse tree, and end up recusing ~10k times because of how deep the parse
> >> tree is. These recursive functions should have a built-in limit. Does
> >> Emacs already have some way to determined the max recursion limit on
> >> each system? Or should we come up with some hard-coded numbers?
> > 
> > Is the recursion in our code, or is it in libtree-sitter?
> 
> In our code, when we walk the parse tree.
> 
> > 
> > If the former, one solution, albeit a crude one, is to track the recursion
> > level and error out if it becomes too deep.  Another solution is to handle
> > the stack in our code, in which case the stack can be allocated on the heap.
> 
> That’s my idea, hence my asking for a reasonable way to get a limit. I think 
> a hard limit is totally reasonable, because there is no way for a “normal” 
> parse tree to be 10k levels deep (that means the source program is 10k levels 
> deep, ver unlikely for any program a human would write or a machine would 
> generated). The one I observed is likely due to the parser misunderstanding 
> the source (due to errors in the code). Plus, I don’t think any user would 
> want to walk that deep into the parse tree either. If someone expects to walk 
> that deep into a parse tree, her program is ill-designed.

How many bytes does each recursive invocation need on the stack?  With that
number at hand, we can estimate a safe value for the limit.  And don't
forget that GC is also highly recursive and eats up a lot of stack space.
I guess GC can happen during building of the imenu list?





reply via email to

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