Re: [PATCH] gdb: consider null terminator for length arguments of value_cstring calls

[Philippe Waroquiers <philippe.waroquiers@skynet.be>]

On Wed, 2021-07-14 at 11:25 +0100, Pedro Alves wrote:
> Such an inferior call in Ada (and other non-terminated string languages) will now
> be passing an unexpected \0 to the called function, of course.  I wonder
> whether that's why Philippe didn't add the \0 in the first place, given his
> Ada bias?  :-)

The idea for $_gdb_setting was to allow a user defined command to access
the gdb settings and provide a behaviour depending on these settings.

I did not envisage the use case of providing the values created by $_gdb_setting
to the inferior.

For the "user defined command" use case, typically, $_streq will be used to 
implement the behaviour depending on the string setting values.

It looks like $_streq works without a null terminator:

(gdb) set var $ccc = "foo"
(gdb) set lang ada
(gdb) set var $aaa := "foo"
(gdb) p $_streq($ccc, $aaa)
$1 = true
(gdb) ptype $ccc
type = array (0 .. 3) of char
(gdb) ptype $aaa
type = array (1 .. 3) of character
(gdb) set lang c
(gdb) p $_streq($ccc, $aaa)
$2 = 1
(gdb) ptype $ccc
type = char [4]
(gdb) ptype $aaa
type = character [3]
(gdb) 

(gdb) set args foo
(gdb) p $_streq($_gdb_setting("args"), "foo")
$1 = 1
(gdb) set lang ada
(gdb) p $_streq($_gdb_setting("args"), "foo")
$2 = true
(gdb) 


So, at least for the initially envisaged use case, I do not see a problem
with the absence (or presence) of a trailing null terminator.

Philippe