Re: [Gnash-dev] Dejagnu parser hell

[Rob Savoye]

strk wrote:
> During an IRC talk we found out that 'make check' is returning different
> results even when called consecutively on a single host/buildtree.

  This is likely due to expect buffering. Expect uses Tcl, but "make
check" is actually an expect program. Any issues with inconsistent
results is usually the pattern that's being looked for doesn't have full
closure, which can cause problems with buffering. Since we're using
expect to grab the output of a Flash movie running under Gnash, there
are probably some weird buffering problems.

> Question is: is there a TCL hacker around willing to make it a
> robust parser, or do we want to implement the parser using a
> bourne shell, or perl ?

  To do so would mean dropping the use of DejaGnu, so as the sole Tcl
hacker, I guess I'll have to see about fixing it.

        - rob -