Re: BUG: Random behavior of the shell

[Sebastian Gniazdowski <sgniazdowski@xxxxxxxxx>]

On Mon, 30 Sep 2019 at 00:34, Bart Schaefer <schaefer@xxxxxxxxxxxxxxxx> wrote:
>
> On Sun, Sep 29, 2019 at 2:03 PM Sebastian Gniazdowski
> <sgniazdowski@xxxxxxxxx> wrote:
> >
> > I was getting such behavior earlier, but waited with the report. Now
> > it's a more clear one, so I'm reporting: https://youtu.be/N-HH2f_Rr0E
>
> It's really difficult to see what's going on with these recordings you
> keep posting.  Video compression often makes them go by even faster
> than real-time, and when the list of results you're displaying more
> than fills the screen it's a pain to step backwards to see what you
> were doing before.  If you must resort to visual aids, a series of
> still screen shots would be better.  That said ...

Ok, I'll remember the tip about the screen shots. The video is a proof
that I'm not making things up, that the shell really does behave in an
unpredictable manner. It also allows to go back and see what was
happening after a time, like when you point later about the long vs.
long long warnings from GCC.

> > The list isn't getting filtered.
> >
> > Earlier was: the `quiet=1` assignment in the below
>
> These don't seem very likely to be closely related problems.  This
> most recent one looks like a pattern matching issue, the other looks
> like perhaps an exit status is leaking from one conditional test to
> another.  Why do you have a [[ ... ]] standing all by itself as a
> command, with nothing testing the results?  (Not that this should
> cause a bug.)

I've minimized the test case, that's why I've removed a print -P from
the conditional command.

I think that the issues are related like when you state later – by a
general problem. It's not the only random behavior that I've observed
– earlier my plugin zconvey just wasn't working, to then start working
without problems. I've now switched to own zsh build and the problems
disappear, so it might be some problem with Mageia's build of Zsh.

> > problems started when I've switched to Linux, so this could not be
> > connected to any recent commits.
>
> This isn't by any chance the same version of linux where Dennis is
> reporting problems with trap functions?

I think not, mine is Mageia 7.

> I saw a "long long" vs. "long" error go by in compiler output in one
> of those videos.  Could wordcode be getting messed up by a default
> integer byte-width in some circumstances, or something like that?

It seems that the warning is related to printf() format string.
However I'm starting to think that this is a problem with a subtle
build error in the Mageia's Zsh distribution. However, shouldn't such
error reveal itself much more often?

-- 
Sebastian Gniazdowski
News: https://twitter.com/ZdharmaI
IRC: https://kiwiirc.com/client/chat.freenode.net:+6697/#zplugin
Blog: http://zdharma.org