Re: [PATCH] Tweaks in gdb.trace/tfile.exp

[Yao Qi <yao at codesourcery dot com>]

On 08/23/2012 01:10 AM, Pedro Alves wrote:
> > @@ -205,9 +205,12 @@ done_making_trace_files (void)
> > >  int
> > >  main (int argc, char **argv, char **envp)
> > >  {
> > >-  write_basic_trace_file ();
> > >+  char *basic_file_name = argv[1];
> > >+  char *error_file_name = argv[2];
> This should have an argc check, for when you run the program
> manually from the command line, with no args, for example.
>
> (I'd have split this change out to that new not yet submitted test you
> mention -- what if review of that new test finds a different approach
> would be better?  No big harm done in this case, so okay.)

This new test need to get a tracefile by executing tfile.exe, but in 
order to avoid the race (this test is using basic.tf while tfile.exp is 
deleting basic.tf), I need the same tracefile with different name. 
Anyway, it will be clear as the new test is submitted.

> > >+# Copy tracefile from target to host.
> > >+remote_download host [remote_upload target b.tf ] basic.tf
> > >+remote_download host [remote_upload target e.tf ] error.tf
> Spurious spaces before ']'.  But, what's going on here?
> The target produces b.tf, and e.tf.  Then those are copied
> to the host with remote_upload.  And then copied again
> to the host, as basic.tf, and error.tf.  Why this juggling?

IIUC, dejagnu is running on machine 'build', so 'remote_upload target 
b.tf' copies b.tf from target to build, instead of host, and then we 
have to use 'remote_download' to copy file from build to host.

--
Yao