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: [RFA v2 00/13] various frame filter fixes and cleanups


On 08/14/2017 04:40 AM, Tom Tromey wrote:

>> So "bt elide" means "elide the elided frames", not "show me the
>> elided frames too".  It's fine with me, though I mildly wonder whether
>> users will be confused by the "double negative".
> 
> "Elide" means "drop", so really "bt elide" should mean "drop whatever
> frames are droppable". Having "bt elide" mean "show the dropped
> frames" would be confusing.  But maybe another word would be better
> here, I don't know.

Maybe "bt no-elided", to go with "bt no-filters".

Because "elided frames" are a preexisting concept, and "elide" is
actually referring to that concept.

Let me try to clarify what I meant.  Note how the patch has the
confusing (to me) "elide the elided" "double negative":

+  if ((flags & PRINT_ELIDE) == 0)
+    {
+      /* Finally recursively print elided frames, if any.  */
+      gdbpy_ref<> elided (get_py_iter_from_func (filter, "elided"));
+      if (elided == NULL)
+	return EXT_LANG_BT_ERROR;

I.e., if "bt elide" is NOT in effect, frames returned by the
"elided" function are printed:

 If PRINT_ELIDE => don't print elided frames
 If not PRINT_ELIDE => print elided frames

I think my brain wants to read PRINT_ELIDE as PRINT_ELIDED, and
then it it becomes really clearly backwards:

  If PRINT_ELIDED => don't print elided frames
  If not PRINT_ELIDED => print elided frames

The "bt elide" switch name confuses me in the same way.  Looking at the
proposed manual change:

 +@item backtrace elide
 +A Python frame filter might decide to ``elide'' some frames.  Normally
 +such elided frames are still printed, but they are indented relative
 +to the filtered frames that cause them to be elided.  The @code{elide}
 +option causes elided frames to not be printed at all.
  @end table

A filter elides a frames.  And then the elide switch elides those
elided frames from view.  So there's two levels of elision, what I
was calling "double negative".

It would seem more natural to me to have a "bt no-elided" switch
meaning "don't print frames marked as elided":

 @item backtrace no-elided
 A Python frame filter might decide to ``elide'' some frames.  Normally
 such elided frames are still printed, but they are indented relative
 to the filtered frames that cause them to be elided.  The @code{no-elided}
 option causes elided frames to not be printed at all.

Or alternatively, to make gdb not print elided frames by default and
add a "bt elided" switch to print them:

 @item backtrace elided
 A Python frame filter might decide to ``elide'' some frames.  Normally
 such elided frames are not printed.  The @code{elided} option causes
 elided frames to be printed, indented relative to the filtered frames
 that cause them to be elided.  

Maybe it's just me and/or a case of "lost in translation", and I'm
fine with that.  Please do feel free to ignore.

Thanks,
Pedro Alves


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