"Chris M. Thomasson" <no@[EMAIL PROTECTED]
> wrote in message
news:kj3Ek.343$YN3.250@[EMAIL PROTECTED]
>I have used it on several in-house algorithms... All is fine, EXCEPT for
1.
>This HARD-CORE "subliminal" little bastard sack of ***** race-condition
was
>sneaking under my nose for YEARS!!!!!!!!!!!!! YEARS!!! OUCH!!! It HURTS.
>
>
>
>
> OH ****%!!!!!
>
>
>
>
> At first I said to myself, potential bug in Relacy wrt Windows emulation
> impl over Fibers... Then I directed Relacys output to a file, printed
it,
> and read the damn thing over and over again. Wow! From that detailed
info,
> I was actually able to create a scenario in which I could create some
> threads, artificially set some key state conditions, and step through
the
> code manually; pausing/resuming threads in the exact places as needed.
> Finally, I could CONSISTENTLY reproduce the racer within the debugger
> using actual _real code_; NOT in the Relacy programming model (e.g., in
> which `$' actually means something! ;^). Well, what a COOL PIECE OF
WORK!
>
> I have to admit, the content of the file that Relacy outputted to was a
> bit overwhelming at first... Then I read it VERY carefully, and over and
> over again; things became crystal clear. My only suggestion, would be to
> create a manual on how to efficiently decipher the error output of
Relacy
> Dmitriy; that would have saved me some time indeed. Perhaps create a
mode
> in which the output can be a little bit more "narrowed/focused"
somewhat.
> Humm... Anyway, GREAT piece of work Dmitriy; seriously!
[...]
Dmitriy, maybe Relacy could create _two_ re****ts...
1. Detailed; narrative style...
2. Action-Items; referring to points within the narrative re****t.
Humm... Does that sound Kosher to you?
|
