This is the mail archive of the gdb-patches@sourceware.org mailing list for the GDB project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [PATCH] Fix -list-thread-groups --available logic and add test


On 10/04/2017 03:32 PM, Simon Marchi wrote:
> On 2017-10-04 10:14, Pedro Alves wrote:
>> On 10/04/2017 03:00 PM, Simon Marchi wrote:
>>
>>> +# List a specific process.
>>> +set spawn_id [remote_spawn target $binfile]
>>> +set pid [spawn_id_get_pid $spawn_id]
>>> +mi_gdb_test \
>>> +    "-list-thread-groups --available i${pid}" \
>>> +    "\\^done,groups=\\\[{id=\"${pid}\",type=\"process\",.*}\\\]" \
>>> +    "list available thread groups with filter"
>>
>> Won't that ".*" above end up matching more than one process, if
>> GDB's filtering ends up a bit borked?  Can we be sure that
>> only one process comes out?
> 
> I tried designing it like that but found it was pretty difficult because
> of the description field:
> 
>  55
> ^done,groups=[{id="499",type="process",description="/home/emaisin/build/binutils-gdb/gdb/testsuite/outputs/gdb.mi/list-thread-groups-available/list-thread-groups-available",user="emaisin",cores=["1"]}]^M
> 
> 
> If we have something like this:
> 
>   groups=[{id="$decimal",type="process",description=".*",...}]

Instead of .*, how about matching "all but unquoted/terminating
double-quote" ?

See http://wiki.tcl.tk/989 for example (Match a "quoted string").

(Maybe there's an easier way to write that regex, haven't
really thought about it.)

> 
> and there are two process entries, the .* in the description field could
> match from the beginning of the first description field to the end of
> the second description field.  So we wouldn't catch that there are two
> processes instead of one.  And we can't really restrict its content,
> because it could potentially include [,}"].
> 
> Do you have any idea on how to do this with a regex?  The only "real"
> solution I see is to have a proper MI parser that parses MI to a tcl
> data structure, and assert that the list of groups is of length 1.

A "real" parser would consume the "..." description string as a single
string by looking for an unescaped terminating double-quote.  :-)

Thanks,
Pedro Alves


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]