Re: [PATCH 2/2] Test case for Bug 28308

[Simon Marchi <simark@simark.ca>]

On 2021-10-01 9:00 p.m., Kevin Buettner via Gdb-patches wrote:
> +# Make sure that there's no running gdb.  It's not clear to me
> +# that this is necessary, but a search shows that all other uses of
> +# gdb_spawn first do a gdb_exit.
> +gdb_exit

I'm not sure why that would be necessary here, as there is nothing above
your gdb_exit that spawns GDB.  If you had used prepare_for_testing
instead of build_executable, then that would have started GDB, and your
gdb_exit would have been necessary (although I would have told you,
don't use prepare_for_testing if you are to exit GDB right after, use
build_executable :)).  So if things work without this gdb_exit (which I
expect), let's just remove it.

> +# Create context in which the global, GDBFLAGS, will be restored at
> +# the end of the block.  All commands run within the block are
> +# actually run in the outer context.  (This is why 'res' is available
> +# outside of the save_vars block.)
> +save_vars { GDBFLAGS } {
> +    # Set flags with which to start GDB.
> +    append GDBFLAGS " -ex \"set height unlimited\""
> +    append GDBFLAGS " -x \"$x_file\""
> +    append GDBFLAGS " --args \"$binfile\""
> +
> +    # Start GDB with above flags.
> +    set res [gdb_spawn]
> +}
> +
> +set test "Load and run script with -x"

Non-capital "l".

> +if { $res != 0} {
> +    fail $test
> +    return -1
> +}
> +
> +# The script loaded via -x contains a run command; while running, GDB
> +# is expected to print three messages from dprintf breakpoints along
> +# with three interspersed messages from an ordinary breakpoint (which
> +# was set up with a continue command).  Set up pattern D to match
> +# output from hitting the dprintf breakpoint and B for the ordinary
> +# breakpoint.  Then set PAT to contain the entire pattern of expected
> +# output from the interspersed dprintf and ordinary breakpoints along
> +# with some (additional) expected output from the dprintf breakpoints,
> +# i.e. 0, 1, and 2.
> +set d "dprintf in increment.., vi="
> +set b "Breakpoint ., inc_vi"
> +set pat "${d}0.*?$b.*?${d}1.*?$b.*?${d}2.*?$b.*?"
> +
> +proc do_test {cmd test} {
> +    gdb_test_multiple $cmd $test {
> +	-re "$::pat$::inferior_exited_re normally.*$::gdb_prompt $" {
> +	    pass $test
> +	}
> +	-re "Don't know how to run.*$::gdb_prompt $" {
> +	    unsupported $test
> +	}

Given you used target_can_use_run_cmd at the top, when would this
unsupported path be taken?

> +    }
> +}
> +
> +# Check output from running script with -x
> +do_test "" $test
> +
> +# Restart GDB and 'source' the script; this will (still) run the program
> +# due to the 'run' command in the script.
> +clean_restart $binfile
> +do_test "source $x_file" "Load and run script using source command"

Non-capital "l".

> +
> +# This should leave us at the gdb prompt; Run program again using
> +# already established breakpoints, i.e. those loaded from the
> +# script.  Prior to fixing PR 28308, this was the only test that
> +# would pass.
> +do_test "run" "Run again"

Non-capital "r".

Simon