Thanks for the hints, Hans. But the problem has been reproduced multiple times after the last update.

The bug turns out to be not from within context. I just tried `touch context-lmtx/tex` (in which there are texmf/, texmf-context/, etc.) and the very same problem appears.

This is very weird, as the whole context-lmtx/ has no more than a few thousand files and <400MB of data, so it's unlikely to be a cache issue either.



On Sun, 12 Sept 2021 at 11:38, Hans Hagen <j.hagen@xs4all.nl> wrote:
On 9/12/2021 1:15 AM, Sylvain Hubert wrote:
> Dear ConTeXt Devs,
>
> Here's a problem that I reported last year
> (https://mailman.ntg.nl/pipermail/ntg-context/2020/099862.html
> <https://mailman.ntg.nl/pipermail/ntg-context/2020/099862.html>) with a
> wrong description. I thought that it had something to do with Firefox or
> browsers, but it turns out to be a much deeper and weirder problem
> messing up with the X window system.
>
> Environment: Manjaro Linux 21.1.2, Xfce 4.16
>
> Steps to reproduce:
> 1. `context --version`  # it can hardly be any more innocent
> 2.1. keep striking random keys in the terminal or mounsepad or whatever
> 2.2. or open a running-cat.gif in the background
> Expected behavior: the characters should appear as you type, or the cat
> should keep running
> Actual behavior: ~3s after `context --version` finishes, the terminal
> freezes and the character stops appearing for ~1s, or similarly, the cat
> stops running for ~1s
>
> In short, ~3s upon termination of context, the X window stuck for ~1s.
>
> More information:
> 1. The problem only appears 2 days after the fresh installation of
> latest context-lmtx, which was also the case last year.
> 2. `context --version` isn't doing anything, so it shouldn't be a
> resource problem, and I've got no such problem keeping 40+ pages open in
> chrome anyway.
> 3. Below is a more precise timing of the problem. The first output is
> the result of `context --version` along with the start and stop
> timestamps. The second output is a python one-liner printing timestamps
> each time it reads an <enter>. Notice how python reads an <enter> per
> ~0.05s until 00:33:21.597326, when it is blocked for ~0.7s.
>
> I'm not sure if it's reproducible everywhere, and I'm willing to debug
> it myself, but I would need the source code of the context executable.
i can't be of much help here but keep i nind that after an update a
first run can trigger a format remake or scan for fonts so when that
takes time hitting some 'run again' key can result in many simulatious
runs that can add up an dmaybe stall the system

Hans


-----------------------------------------------------------------
                                           Hans Hagen | PRAGMA ADE
               Ridderstraat 27 | 8061 GH Hasselt | The Netherlands
        tel: 038 477 53 69 | www.pragma-ade.nl | www.pragma-pod.nl
-----------------------------------------------------------------