Re: Separating stdout and stderr

On 26 March 2016 at 00:25, Ben Elliston <[email protected]> wrote:
> On Fri, Mar 25, 2016 at 05:54:53PM +0100, Christophe Lyon wrote:
>
>> At this point, I don't know where the problem is. We've been seeing
>> random results mostly in the sanitizer tests for months.
>
> I'll try and look into this.
>

Thanks.

In the mean time maybe you can help me undestand how I can make
progress on my side?

For instance, I added a dumb:
set outp "/tmp/foo.[pid]"
right before
if { "$inp" == "" && "$outp" == "" }
at the beginning of local_exec (in remote.exp)

I expected this to make several tests fails because their output was redirected.

But what happens is that I get:
ERROR: bad spawn_id (process died earlier?)
while executing
"expect {
     -i $spawn_id -timeout $timeout -re ".+" {
      append output $expect_out(buffer)
      if { [string length $output] < 512000 } {
      exp_contin..."

Maybe it's a dumb question, but I don't understand why forcing redirection
causes an error in tcl/expect, rather than at the testcase level?

This is happening while running the binutils tests (i.e. well before starting
the gcc tests)

Christophe