RE: [cdt-patch] Debug timeout and cleanup patch.

[Thomas Fletcher <ThomasF@xxxxxxx>]

Incidentally there are related PRs on this already: 
  PR 39758 - Request for a change in the timeout behaviour (improved by this
patch).
  PR 43496 - Describes the problem of having gdb processes not properly
terminated.

This patch helps the situation of both.

Thomas

> -----Original Message-----
> From: Thomas Fletcher [mailto:ThomasF@xxxxxxx]
> Sent: September 22, 2003 12:16 PM
> To: 'cdt-patch@xxxxxxxxxxx'
> Subject: [cdt-patch] Debug timeout and cleanup patch.
> 
> 
> 
> There are a number of problems we have when launching applications
> which have a lot of symbols.  Often times if during that launch
> period (since we provide the executable on the command line to gdb
> and gdb will slurp up the symbol information automatically) if your
> timeout value is too low, then we will throw and exception and kill
> the session.  This is bad for a couple of reasons:
> 
> 1- We should be able to do some sort of intelligent heuristic when
>    it comes to scaling these types of things.
> 
> This is a prototype of a "general fix" which looks at the executable
> size to generate an approximated timeout and then compares 
> the session 
> timeout with that size.  If the approximated timeout is greater than
> the user defined session timeout it will set the session timeout to
> a longer value.
> 
> 2- Even when things go wrong, we should make sure that we clean up
>    properly after ourselves ... and if there is no session created,
>    this means cleaning up the gdb process we created.
> 
> This fix catches errors in the session creation and kills off the
> process before re-throwing the exception.
> 
> Thanks,
>  Thomas
> 
>