TR: ARI warning list change Fri May 29 01:53:26 UTC 2009

TR: ARI warning list change Fri May 29 01:53:26 UTC 2009

[Pierre Muller]

Here is another example of the email that is generated
by my patch to update-web-ari.

  It seems like you have forgotten a bunch of _() 
macros.

  A possible improvement would be to send this message directly to
people that committed a change to the sources that
generated a new warning.

  Is it possible to extract the commits
from the last 24 hours to find out to whom this kind of message should be
sent?

Pierre Muller
Still trying to improve ARI.


-----Message d'origine-----
De : GDB Administrator [mailto:gdbadmin@sourceware.org] 
Envoyé : Friday, May 29, 2009 3:53 AM
À : muller@sourceware.org
Objet : ARI warning list change Fri May 29 01:53:26 UTC 2009

657a658,659
> gdb/mi/mi-cmd-var.c:251: gettext: _ markup: All messages should be marked
up with _.
> gdb/mi/mi-cmd-var.c:256: gettext: _ markup: All messages should be marked
up with _.
837a840,846
> gdb/varobj.c:853: gettext: _ markup: All messages should be marked up with
_.
> gdb/varobj.c:859: gettext: _ markup: All messages should be marked up with
_.
> gdb/varobj.c:865: gettext: _ markup: All messages should be marked up with
_.
> gdb/varobj.c:882: gettext: _ markup: All messages should be marked up with
_.
> gdb/varobj.c:1386: gettext: _ markup: All messages should be marked up
with _.
> gdb/varobj.c:1457: gettext: _ markup: All messages should be marked up
with _.
> gdb/varobj.c:1470: gettext: _ markup: All messages should be marked up
with _.
1204d1212
< gdb/value.c:529: deprecated: deprecated value_address_hack: Do not use
deprecated value_address_hack, see declaration for details
1395,1396d1402
< gdb/value.h:278: deprecated: deprecated value_address_hack: Do not use
deprecated value_address_hack, see declaration for details
< gdb/value.h:279: deprecated: deprecated value_address_hack: Do not use
deprecated value_address_hack, see declaration for details

Re: TR: ARI warning list change Fri May 29 01:53:26 UTC 2009

[Vladimir Prus]

On Friday 29 May 2009 10:52:19 Pierre Muller wrote:
> Here is another example of the email that is generated
> by my patch to update-web-ari.
> 
>   It seems like you have forgotten a bunch of _() 
> macros.
> 
>   A possible improvement would be to send this message directly to
> people that committed a change to the sources that
> generated a new warning.
> 
>   Is it possible to extract the commits
> from the last 24 hours to find out to whom this kind of message should be
> sent?
> 
> Pierre Muller
> Still trying to improve ARI.
> 
> 
> -----Message d'origine-----
> De : GDB Administrator [mailto:gdbadmin@sourceware.org] 
> Envoyé : Friday, May 29, 2009 3:53 AM
> À : muller@sourceware.org
> Objet : ARI warning list change Fri May 29 01:53:26 UTC 2009
> 
> 657a658,659
> > gdb/mi/mi-cmd-var.c:251: gettext: _ markup: All messages should be marked
> up with _.
> > gdb/mi/mi-cmd-var.c:256: gettext: _ markup: All messages should be marked
> up with _.
> 837a840,846
> > gdb/varobj.c:853: gettext: _ markup: All messages should be marked up with
> _.
> > gdb/varobj.c:859: gettext: _ markup: All messages should be marked up with
> _.
> > gdb/varobj.c:865: gettext: _ markup: All messages should be marked up with
> _.
> > gdb/varobj.c:882: gettext: _ markup: All messages should be marked up with
> _.
> > gdb/varobj.c:1386: gettext: _ markup: All messages should be marked up
> with _.
> > gdb/varobj.c:1457: gettext: _ markup: All messages should be marked up
> with _.
> > gdb/varobj.c:1470: gettext: _ markup: All messages should be marked up
> with _.
> 1204d1212
> < gdb/value.c:529: deprecated: deprecated value_address_hack: Do not use
> deprecated value_address_hack, see declaration for details
> 1395,1396d1402
> < gdb/value.h:278: deprecated: deprecated value_address_hack: Do not use
> deprecated value_address_hack, see declaration for details
> < gdb/value.h:279: deprecated: deprecated value_address_hack: Do not use
> deprecated value_address_hack, see declaration for details

I am missing context on this email, but at least in MI, it's not yet clear
which messages should be i18n and which are not. Some are results of frontend
coding bug and as such should be never presented to user and should not be
ever translated. And if they are translated, they won't help the user but
will make it harder to understand the error for developer.

- Volodya

RE: TR: ARI warning list change Fri May 29 01:53:26 UTC 2009

[Pierre Muller]

> > Objet : ARI warning list change Fri May 29 01:53:26 UTC 2009
> >
> > 657a658,659
> > > gdb/mi/mi-cmd-var.c:251: gettext: _ markup: All messages should be
> marked

> I am missing context on this email, but at least in MI, it's not yet
> clear
> which messages should be i18n and which are not. Some are results of
> frontend
> coding bug and as such should be never presented to user and should not
> be
> ever translated. And if they are translated, they won't help the user
> but
> will make it harder to understand the error for developer.


  You are probably right... I don't know much about mi...

  Should this warning disappear for mi directory altogether,
or should we add a /* ARI: _ markup */ to silence gdb_ari.sh
where translation is discouraged?

Pierre

PS: Could you take care of the missing "_()"
in gdb/varobj.c source?

    I think that these are the kind of patches that can go in
as obvious, as all mechanical ARI fixes.
    I hope that I translated correctly here my
remembering of discussions that we had about this.

Re: TR: ARI warning list change Fri May 29 01:53:26 UTC 2009

[Vladimir Prus]

On Friday 29 May 2009 12:31:30 Pierre Muller wrote:
> > > Objet : ARI warning list change Fri May 29 01:53:26 UTC 2009
> > >
> > > 657a658,659
> > > > gdb/mi/mi-cmd-var.c:251: gettext: _ markup: All messages should be
> > marked
> 
> > I am missing context on this email, but at least in MI, it's not yet
> > clear
> > which messages should be i18n and which are not. Some are results of
> > frontend
> > coding bug and as such should be never presented to user and should not
> > be
> > ever translated. And if they are translated, they won't help the user
> > but
> > will make it harder to understand the error for developer.
> 
> 
>   You are probably right... I don't know much about mi...
> 
>   Should this warning disappear for mi directory altogether,
> or should we add a /* ARI: _ markup */ to silence gdb_ari.sh
> where translation is discouraged?
> 
> Pierre
> 
> PS: Could you take care of the missing "_()"
> in gdb/varobj.c source?

I don't really know what is "missing". Am I supposed to
grep for such, or there's a list of suspect places already?

> 
>     I think that these are the kind of patches that can go in
> as obvious, as all mechanical ARI fixes.

I *clearly* miss some context. What is "ARI"? What is this automated
inspection thing? Where are results of such inspection posted?

- Volodya

RE: TR: ARI warning list change Fri May 29 01:53:26 UTC 2009

[Pierre Muller]

Sorry,

The ARI stands for Awk Regression index
and is a script (named gdb_ari.sh) that is run once a day by
the machine on which the webite is.

The result of this script is transformed into a web page:
http://sourceware.org/gdb/current/ari/

Most used files are accessible at the bottom of the web page,
in particular gdb_ari.sh is at:
http://sourceware.org/gdb/current/ari/gdb_ari.sh


A recent change I made adds a email generation that is
only sent to me until now, but I was trying to
get this sent to gdb-testers.
The email message can be seen of the website:
http://sourceware.org/gdb/current/ari/mail-message

This message shows the differences between
yesterday's and today's list of warnings generated by gdb_ari.sh.


The email that I got this night was
>>>>>>>>>>Current version of email message
657a658,659
> gdb/mi/mi-cmd-var.c:251: gettext: _ markup: All messages should be marked
up with _.
> gdb/mi/mi-cmd-var.c:256: gettext: _ markup: All messages should be marked
up with _.
837a840,846
> gdb/varobj.c:853: gettext: _ markup: All messages should be marked up with
_.
> gdb/varobj.c:859: gettext: _ markup: All messages should be marked up with
_.
> gdb/varobj.c:865: gettext: _ markup: All messages should be marked up with
_.
> gdb/varobj.c:882: gettext: _ markup: All messages should be marked up with
_.
> gdb/varobj.c:1386: gettext: _ markup: All messages should be marked up
with _.
> gdb/varobj.c:1457: gettext: _ markup: All messages should be marked up
with _.
> gdb/varobj.c:1470: gettext: _ markup: All messages should be marked up
with _.
1204d1212
< gdb/value.c:529: deprecated: deprecated value_address_hack: Do not use
deprecated value_address_hack, see declaration for details
1395,1396d1402
< gdb/value.h:278: deprecated: deprecated value_address_hack: Do not use
deprecated value_address_hack, see declaration for details
< gdb/value.h:279: deprecated: deprecated value_address_hack: Do not use
deprecated value_address_hack, see declaration for details
>>>>>>>>>>End of current version of email message

This shows that your patch generated new warnings,
but also removed three warnings about the deprecated use of
value_address_hack
that is completely removed. This is great.

  But it also lists several positions,
with file source and line number of
additional warnings. For that patch all new warnings
are of the same kind. This is one of the rules checked
by gdb_ari.sh that verifies that all constant messages
are enclosed in _() markers for internalization.

  As you pointed out, the mi code might be unsuitable
for these markers, but I suppose that for the gdb/varobj.c

  A newer version of update-web-ari (that I didn't commit yet)
improves the email by also listing the corresponding source line:
>>>>>>>>>>New version of email message
941a942,943
> gdb/mi/mi-cmd-var.c:251: gettext: _ markup: All messages should be marked
up with _.
gdb/mi/mi-cmd-var.c:251:    error ("Usage: NAME VISUALIZER_FUNCTION.");
> gdb/mi/mi-cmd-var.c:256: gettext: _ markup: All messages should be marked
up with _.
gdb/mi/mi-cmd-var.c:256:    error ("Variable object not found");
1724d1725
< gdb/value.c:529: deprecated: deprecated value_address_hack: Do not use
deprecated value_address_hack, see declaration for details
gdb/value.c:529:deprecated_value_address_hack (struct value *value)
1735,1736d1735
< gdb/value.h:278: deprecated: deprecated value_address_hack: Do not use
deprecated value_address_hack, see declaration for details
gdb/value.h:278:extern CORE_ADDR *deprecated_value_address_hack (struct
value *);
< gdb/value.h:279: deprecated: deprecated value_address_hack: Do not use
deprecated value_address_hack, see declaration for details
gdb/value.h:279:#define VALUE_ADDRESS(val) (*deprecated_value_address_hack
(val))
1742a1742,1748
> gdb/varobj.c:853: gettext: _ markup: All messages should be marked up with
_.
gdb/varobj.c:853:      error ("Null value returned for children");
> gdb/varobj.c:859: gettext: _ markup: All messages should be marked up with
_.
gdb/varobj.c:859:    error ("Returned value is not iterable");
> gdb/varobj.c:865: gettext: _ markup: All messages should be marked up with
_.
gdb/varobj.c:865:      error ("Could not get children iterator");
> gdb/varobj.c:882: gettext: _ markup: All messages should be marked up with
_.
gdb/varobj.c:882:	error ("Invalid item from the child list");
> gdb/varobj.c:1386: gettext: _ markup: All messages should be marked up
with _.
gdb/varobj.c:1386:  error ("Python support required");
> gdb/varobj.c:1457: gettext: _ markup: All messages should be marked up
with _.
gdb/varobj.c:1457:      error ("Could not evaluate visualizer expression:
%s", visualizer);
> gdb/varobj.c:1470: gettext: _ markup: All messages should be marked up
with _.
gdb/varobj.c:1470:  error ("Python support required");
>>>>>>>>>>End of new version of email message 
  
  By looking at this I tend to think that these messages
should indeed be translated if we ever translate gdb...

 I hope this explains a little bit what ARI is all about.

Pierre

Re: TR: ARI warning list change Fri May 29 01:53:26 UTC 2009

[Thiago Jung Bauermann]

Hi Pierre,

Em Sexta-feira 29 Maio 2009, às 06:42:22, Pierre Muller escreveu:
> The ARI stands for Awk Regression index
> and is a script (named gdb_ari.sh) that is run once a day by
> the machine on which the webite is.
>
> The result of this script is transformed into a web page:
> http://sourceware.org/gdb/current/ari/
>
> Most used files are accessible at the bottom of the web page,
> in particular gdb_ari.sh is at:
> http://sourceware.org/gdb/current/ari/gdb_ari.sh

Thanks for the info, I always wondered myself what the ARI meant. :-)

Just added the text above to the GDB wiki:

http://sourceware.org/gdb/wiki/DeveloperTips

-- 
[]'s
Thiago Jung Bauermann
Software Engineer
IBM Linux Technology Center

Re: TR: ARI warning list change Fri May 29 01:53:26 UTC 2009

[Tom Tromey]

>>>>> "Pierre" == Pierre Muller <muller@ics.u-strasbg.fr> writes:

Pierre> Here is another example of the email that is generated
Pierre> by my patch to update-web-ari.

Pierre>   It seems like you have forgotten a bunch of _() 
Pierre> macros.

Sorry about that.  I am going to commit the appended patch to fix the
problem.

Pierre>   A possible improvement would be to send this message directly to
Pierre> people that committed a change to the sources that
Pierre> generated a new warning.

Pierre>   Is it possible to extract the commits from the last 24 hours
Pierre> to find out to whom this kind of message should be sent?

With CVS, it can be done, but very easily.

Tom

2009-06-05  Tom Tromey  <tromey@redhat.com>

	* varobj.c (update_dynamic_varobj_children): Wrap error text in
	_().
	(install_visualizer): Likewise.
	(varobj_set_visualizer): Likewise.

Index: varobj.c
===================================================================
RCS file: /cvs/src/src/gdb/varobj.c,v
retrieving revision 1.132
diff -u -r1.132 varobj.c
--- varobj.c	4 Jun 2009 21:56:22 -0000	1.132
+++ varobj.c	5 Jun 2009 18:49:57 -0000
@@ -850,19 +850,19 @@
   if (!children)
     {
       gdbpy_print_stack ();
-      error ("Null value returned for children");
+      error (_("Null value returned for children"));
     }
 
   make_cleanup_py_decref (children);
 
   if (!PyIter_Check (children))
-    error ("Returned value is not iterable");
+    error (_("Returned value is not iterable"));
 
   iterator = PyObject_GetIter (children);
   if (!iterator)
     {
       gdbpy_print_stack ();
-      error ("Could not get children iterator");
+      error (_("Could not get children iterator"));
     }
   make_cleanup_py_decref (iterator);
 
@@ -879,7 +879,7 @@
       inner = make_cleanup_py_decref (item);
 
       if (!PyArg_ParseTuple (item, "sO", &name, &py_v))
-	error ("Invalid item from the child list");
+	error (_("Invalid item from the child list"));
       
       if (PyObject_TypeCheck (py_v, &value_object_type))
 	{
@@ -1383,7 +1383,7 @@
   if (!visualizer && var->children_requested)
     varobj_list_children (var);
 #else
-  error ("Python support required");
+  error (_("Python support required"));
 #endif
 }
 
@@ -1454,7 +1454,7 @@
   if (! pretty_printer)
     {
       gdbpy_print_stack ();
-      error ("Could not evaluate visualizer expression: %s", visualizer);
+      error (_("Could not evaluate visualizer expression: %s"), visualizer);
     }
 
   if (pretty_printer == Py_None)
@@ -1467,7 +1467,7 @@
 
   do_cleanups (back_to);
 #else
-  error ("Python support required");
+  error (_("Python support required"));
 #endif
 }
 

Re: TR: ARI warning list change Fri May 29 01:53:26 UTC 2009

[Tom Tromey]

>>>>> "Tom" == Tom Tromey <tromey@redhat.com> writes:

Pierre> Is it possible to extract the commits from the last 24 hours
Pierre> to find out to whom this kind of message should be sent?

Tom> With CVS, it can be done, but very easily.

Of course, I meant to say, *not* very easily.

Could we run some kind of ARI regression check as part of
"make check"?

Tom