Re: [PATCH v2] Raise exception if ambiguous name is used in gdb.parameter

[Hannes Domani <ssbssa@yahoo.de>]

 Am Montag, 29. Januar 2024 um 14:19:23 MEZ hat Andrew Burgess <aburgess@redhat.com> Folgendes geschrieben:

> Hannes Domani <ssbssa@yahoo.de> writes:
>
> > Currently gdb.parameter doesn't raise an exception if an
> > ambiguous name is used, it instead returns the value of the
> > last partly matching parameter:
> > ```
> > (gdb) show print sym
> > Ambiguous show print command "sym": symbol, symbol-filename, symbol-loading.
> > (gdb) show print symbol-loading
> > Printing of symbol loading messages is "full".
> > (gdb) py print(gdb.parameter("print sym"))
> > full
> > ```
> >
> > It's because lookup_cmd_composition_1 tries to detect
> > ambigous names by checking the return value of find_cmd
> > for CMD_LIST_AMBIGUOUS, which never happens, since only
> > lookup_cmd_1 returns CMD_LIST_AMBIGUOUS.
> > Instead the nfound argument contains the number of found
> > matches.
> >
> > By using it instead, and by setting *CMD to the special value
> > CMD_LIST_AMBIGUOUS in this case, gdbpy_parameter can now show
> > the appropriate error message:
> > ```
> > (gdb) py print(gdb.parameter("print sym"))
> > Traceback (most recent call last):
> >  File "<string>", line 1, in <module>
> > RuntimeError: Parameter `print sym' is ambiguous.
> > Error while executing Python code.
> > (gdb) py print(gdb.parameter("print symbol"))
> > True
> > (gdb) py print(gdb.parameter("print symbol-"))
> > Traceback (most recent call last):
> >  File "<string>", line 1, in <module>
> > RuntimeError: Parameter `print symbol-' is ambiguous.
> > Error while executing Python code.
> > (gdb) py print(gdb.parameter("print symbol-load"))
> > full
> > ```
> >
> > Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=14639
> > ---
> > v2:
> > - Small changes in commit message.
> > - Use CMD_LIST_AMBIGUOUS in lookup_cmd_composition_1 to mark
> >  ambiguous commands, so gdbpy_parameter can show a better
> >  error message.
> > ---
> >  gdb/cli/cli-decode.c                      | 13 ++++++---
> >  gdb/python/python.c                      |  5 +++-
> >  gdb/testsuite/gdb.python/py-parameter.exp | 32 +++++++++++++++++++++++
> >  3 files changed, 46 insertions(+), 4 deletions(-)
> >
> > diff --git a/gdb/cli/cli-decode.c b/gdb/cli/cli-decode.c
> > index 940cd6a2c8e..db57764578d 100644
> > --- a/gdb/cli/cli-decode.c
> > +++ b/gdb/cli/cli-decode.c
> > @@ -2549,7 +2549,8 @@ deprecated_cmd_warning (const char *text, struct cmd_list_element *list)
> >    If TEXT is a subcommand (i.e. one that is preceded by a prefix
> >    command) set *PREFIX_CMD.
> >
> > -  Set *CMD to point to the command TEXT indicates.
> > +  Set *CMD to point to the command TEXT indicates, or to
> > +  CMD_LIST_AMBIGUOUS if there are multiple possible matches.
> >
> >    If any of *ALIAS, *PREFIX_CMD, or *CMD cannot be determined or do not
> >    exist, they are NULL when we return.
> > @@ -2589,7 +2590,12 @@ lookup_cmd_composition_1 (const char *text,
> >        *cmd = find_cmd (command.c_str (), len, cur_list, 1, &nfound);
> >
> >        /* We only handle the case where a single command was found.  */
> > -      if (*cmd == CMD_LIST_AMBIGUOUS || *cmd == nullptr)
> > +      if (nfound > 1)
> > +    {
> > +      *cmd = CMD_LIST_AMBIGUOUS;
> > +      return 0;
>
> This changes the API of lookup_cmd_composition.  You need to audit all
> the calls to this function to check that non are depending on CMD being
> set to nullptr.  I only looked at one in cli/cli-script.c in the
> function do_document_command that is a problem, try this:
>
>   (gdb) define foo_1
>   Type commands for definition of "foo_1".
>   End with a line saying just "end".
>   >echo aa\n
>   >end
>   (gdb) define foo_2
>   Type commands for definition of "foo_2".
>   End with a line saying just "end".
>   >echo bb\n
>   >end
>   (gdb) document foo
>  
>  
>   Fatal signal: Segmentation fault
>   ... etc ...
>
> Not every use of lookup_cmd_composition is a problem, e.g. in
> alias_command in cli/cli-cmds.c it's not an issue, so you'll need to
> check 'em all.
>
> If you do find cases that would have triggered bugs with this change, it
> would be awesome if you could add tests to cover these cases, though I
> guess that's not a hard requirement.

Thanks for catching this.
I've now checked all uses of lookup_cmd_composition, and could not find
another problem (either it bailed because of lookup_cmd_composition return
value, or the returned *cmd was not used at all afterwards).

So I've fixed do_document_command, and added a test for this, in the v3:
https://sourceware.org/pipermail/gdb-patches/2024-January/206257.html


Hannes