bug#10652: font-lock very slow for C++

[Toon Claes]

Hi Alan,

 

emacs hangs for hours. Luckily it only used one of my 4 processor cores, so I could leave it loading in the background.

This is what I've done:

Open the file, in c++-mode, and scroll to end-of-buffer (M->). The mouse cursor starts spinning, but the end of buffer is not shown.

When I press C-g three times, the end of the buffer is shown. The fontification looks correct.

When I go back to beginning-of-buffer, same problem.

When I press M->, M-< several times, I cannot recover with C-g and I need to kill emacs:-(

 

isearch works fine with font-lock-mode disabled.

Even, depending on the position in the buffer, it happens the command M-x font-lock-mode gives the spinning mouse cursor. Again 3x C-g solves this.

So I can say pretty sure, font-lock-mode is causing the "delay".

 

I've tested again with the bazaar version of last friday.

 

My test file contains something like this:

    int ClassName::MethodAbc(void)

    {

      SOME_PRETEST_MACRO;

      return TranslateResult(LibFunctionAbc(m_Member));

    }

 

And this repeated for something like 20 times (of course with different names, and with different parameters).

The macro is something like:

    #define SOME_PRETEST_MACRO   if (!Ready()) return -1;

 

With this test file, I could easily reproduce the problem.

I hope this can help investigating the issue.

 

 

Toon

 

 

 

On 2012-03-30 00:00, Alan Mackenzie wrote:

> Hello, Toon.
>
> On Thu, Mar 29, 2012 at 09:46:03PM +0200, Toon Claes wrote:
> > Hi Alan,
> > About: (i) yes I've tried that, emacs -Q didn't make a difference
> Worth knowing
> > (ii) I am using emacs from git, commit 2ac37884107bd4e78bb… Should I get it from the bazaar repo?
> I'm not familiar with the git repository.  If the revision is recent
> (within the last week or two or three) it should be OK.  I just wanted to
> check you weren't using a 6-month old version.
> > (iii) I know font-lock has nothing to do with isearch. When I disable font-lock-mode for the buffer, isearch does not hang. So the issue isn't isearch. Sorry I created some confusion in previous mail.
> OK.  What I suspect is that the isearch works fine, then the
> fontification hangs before the screen can redisplay.  How long does it
> hang for?  Any chance you could perhaps leave it hanging over lunchtime,
> or perhaps even overnight to see if it might just be _very_ slow.
>
> Could you perhaps try disabling font-lock, doing the isearch, then
> enabling font-lock again.  Does it then still hang, or does it come up
> OK?
>
> Lastly, is there anything unusual about your test file.  Perhaps long
> regions of text which contain no semicolons or maybe no braces?  Or maybe
> hundreds of macro definitions one after the other, something like that?
> > Toon