lmi
[Top][All Lists]
Advanced

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

Re: [lmi] Can wx_test say "acdgimprsxcciim"?


From: Greg Chicares
Subject: Re: [lmi] Can wx_test say "acdgimprsxcciim"?
Date: Wed, 05 Nov 2014 01:54:45 +0000
User-agent: Mozilla/5.0 (Windows NT 5.1; rv:24.0) Gecko/20100101 Thunderbird/24.6.0

On 2014-11-05 00:22Z, Vadim Zeitlin wrote:
> On Tue, 04 Nov 2014 16:07:05 +0000 Greg Chicares <address@hidden> wrote:
> 
> GC> An odd thing happened to me today.
> 
> [For the record, I had sent my private Halloween email to you before
>  reading this. Things are definitely getting spookier and spookier.
>  I advise you to update all your backups.]

Can't--the backup drive is in a ten-year-old computer that no longer
recognizes the drive, and whose uninterruptible power supply recently
failed. Just in case that makes you feel any better.

> GC> I started rebuilding wx from scratch, and it occurred to me that it
> GC> might be a good idea to run 'wx_test' one last time before upgrading
> GC> wx--so I ran it right after I started the wx rebuild, while wx was
> GC> still running its 'configure' step.
> GC> 
> GC> All 24 tests failed. The font changed in the separate zsh session
> GC> where I was running 'wx_test'. And "acdgimprsxcciim%" appeared in
> GC> that zsh session. The "%" was highlighted, indicating the absence
> GC> of a newline. But where did "acdgimprsxcciim" come from? Could
> GC> that be simulated keyboard events from 'wx_test'? I can't explain
> GC> the alphabetical "a" to "x" part, but "cciim" looks suspiciously
> GC> like census-census-illustration-illustration-mec_testing.
> 
>  If the focus has somehow switches to the window containing zsh, it might
> have indeed come from the test. "a" is used in wx_test_about_dialog.cpp,
> which runs first. The next one, wx_test_benchmark_census.cpp, is skipped on
> your machine because of the missing wx_test.conf IIUC. The "cd" must come
> from the next one, wx_test_calculation_summary.cpp. Or maybe they're
> already from the beginning of wx_test_create_open.cpp. In any case,
> "gimprsx" do come from it (GPT, illustration, MEC, policy, rounding,
> strata, text, respectively). And I think your explanation for the last part
> is correct too.

Ah, that's welcome news.

>  Why has the focus switched to the terminal is another question but it
> could have been confused by the firewall prompt. I'm not sure how exactly
> is it implemented as I almost never use Windows firewall

Not that thing--a different firewall.

> myself, but I
> believe it does have the special dispensation to steal focus from the
> active foreground application (which shouldn't normally be possible).

Well, who knows how they do it, but that's how it behaves.

> GC> Furthermore, my firewall told me that 'basename' wanted to
> GC> execute 'cc1'. I can understand why wx configuration would use
> GC> both, but I don't see why 'basename' would invoke an "internal"
> GC> part of gcc. I allowed it, and wx seems to have built correctly.
> GC> But the firewall never previously prompted me to allow this.
> 
>  Sorry, I have absolutely no explanation for this.

Okay, it's a freak thing. I recently upgraded Cygwin, so I'm still
telling the firewall to allow stuff that I don't frequently do with
the new files.

> For one, I have no idea
> whatsoever why would a firewall be monitoring the local program execution
> anyhow. To the best of my knowledge, neither "basename" nor "cc1" use
> network, so how is it involved in the first place?

I guess the definition of "firewall" differs depending on who writes
it. The one I'm using monitors any program starting another, and I've
set it to ask my permission first, but remember my answers--until I
replace Cygwin, and then I have to retrain it. In practice, this is
not a significant burden.

>  Are there any more details about what happened in Windows Event Viewer? If
> building wxWidgets results in firewall warnings, I'd like to know more
> about this...

"Application": nothing in the past year except for the mysterious
event 0 from Cygwin (which last occurred on Sunday) and a few
apparently spurious warnings from my supermicro monitor program
about low battery voltage (which occur sporadically even if I
replace the battery with a fresh one).

"System": nothing in the past month except bad blocks on one old
CD, a few internet exploder crashes (I have to use it to access
company email), and a single TCP/IP warning 4226 from a few days
ago (which presumably is insignificant because it has occurred
only twice this year, months apart).

"Security": nothing at all.

I had built wx from scratch several times already with this new
Cygwin installation, without seeing this basename-calls-cc1
thing. I suppose it's nothing to do with wx in particular; it's
probably just that building wx does run quite a few processes
on sixteen cores (eight physical, doubled for "hyperthreading"),
and presumably the firewall just got confused.




reply via email to

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