strange tcl dejagnu error

strange tcl dejagnu error

[Andrew Cagney]

Hello,

Just tried running a cross testsuite and am seeing (powerpc-netbsd X 
mips-elf):

Running /home/scratch/GDB/src/gdb/testsuite/gdb.base/arithmet.exp ...
ERROR: tcl error sourcing 
/home/scratch/GDB/src/gdb/testsuite/gdb.base/arithmet.exp.
ERROR: expected boolean value but got ""
     while executing
"if [gdb_file_cmd $arg] then { return -1 }"
     (procedure "gdb_load" line 9)
     invoked from within
"gdb_load ${binfile}"
     (file "/home/scratch/GDB/src/gdb/testsuite/gdb.base/arithmet.exp" 
line 50)
     invoked from within
"source /home/scratch/GDB/src/gdb/testsuite/gdb.base/arithmet.exp"
     ("uplevel" body line 1)
     invoked from within
"uplevel #0 source 
/home/scratch/GDB/src/gdb/testsuite/gdb.base/arithmet.exp"
     invoked from within
"catch "uplevel #0 source $test_file_name""

anyone else see this?

Andrew

Re: strange tcl dejagnu error

[Michael Chastain]

Oh, shit.  This is mine.

I changed the return type of gdb_file_cmd; and I did a careful job of
changing all the callers in gdb.*/*.exp and lib/*.exp, but since I test
native only, I did not even think about config/*.exp.

The right code is:

  set status [gdb_file_cmd $arg]
  if { [lindex $status 0] != "" } then {
    # gdb_file_cmd failed
    whatever;
  }

See the comment at the front of gdb_file_cmd for the specification
of the return value.

I'll fix this tomorrow.  Sorry!

(And man, is there a lot of funky dead code in config/*.exp).

Michael

Re: strange tcl dejagnu error

[Michael Chastain]

Andrew Cagney wrote:
> ERROR: expected boolean value but got ""
>      while executing
> "if [gdb_file_cmd $arg] then { return -1 }"
>      (procedure "gdb_load" line 9)
>      invoked from within
> "gdb_load ${binfile}"
>      (file "/home/scratch/GDB/src/gdb/testsuite/gdb.base/arithmet.exp" 

Just so you know, I'm still chewing on this problem.

Plan #1

  I back out the changes from Manoj and myself to make gdb_file_cmd
  return more information so that the caller can tell if the executable
  was stripped or not.  Then all the alternate implementations of
  gdb_load in config/*.exp ought to work again.  After that, I figure
  out some other mechanism to get what we wanted in the first place.

Plan #2

  I go forward, patching up gdb_load in config/*.exp.  I actually
  wrote this patch, but it's 500 lines long, scattered over a lot
  of places, and will be hard to test all of it.  (monitor, gdbserver,
  netware, sid, sim, and so on).

Both plans are icky in different ways.

Gotta stare at the wall some more.  Ideas welcomed.

Michael