Notes on i18n

Notes on i18n

[Philipp Thomas]

As the 2.95 branch has happened, it's time to unearth ones todo lists. And
here is part of mine (other parts will follow in separate posts).

Before I start, a warning for the casual reader: this *will* get rather
long, so if you're not actively taking part in egcs's development,
you may safely skip this message.

	
While working on completing i18n for egcs I've made some observations.
Some of them are not always obvious for those whose native tongue is English,
others I'll mention just to make it simpler for those that haven't dealt
with i18n matters before.

I'd like to share them all with you fellow egcs developers in order to ease my
work (well, wouldn't you have guessed :) and that of other future translators.

In the following I'll abbreviate gcc/ABOUT-GCC-NLS as AGN.

 - Before you implement any diagnostic/warning/error reporting functions,
   please read AGN for a description of the mechanisms used, and for
   guidelines as to what should and what shouldn't be translated.

 - I'd also recommend browsing gettext.texi - part of the GNU gettext
   package - specifically the topics under `Preparing Program Sources' in
   the toplevel menu.

 - if you create a new file, or change an existing one that contains/will
   contain something that should be translated, please update
   gcc/po/POTFILES.in if necessary.

 - In functions that are part of the reporting mechanism of the
   frontend (f.i. cp_error for g++), please name the parameter in which the
   message string is passed msgid or end its name with msgid. gcc/exgettext,
   gcc's helper script searches for these names to determine which strings
   to extract for translation.
   
 - When you create/use static message strings, encapsulate them either with
   N_() if the string will get translated on a higher level, or with _()
   if the string is used locally to construct a larger message.
   This makes the strings visible for the message extraction tool(s).
   Here an example from java/parse.c (out of my to_send queue :) :

      char *msg = (!valid_cast_to_p (type_value, type) ?
		   _("Can't") : _("Explicit cast needed to"));

 - If you must use printf et al, mark the message with _(). Again, please see
   to it that direct printfs are concentrated in as few routines as possible
   in order to avoid cluttering up the sources with _().

 - When using conditional expressions for printing the plural form, _please_
   use the full words, rather than just the endings, e.g. 
   
     printf ("%d %s\n", n, (n == 1 ? "warning" : "warnings");
   
   instead of 
   
     printf("%d warning%s\n", n, (n == 1 ? "" : "s")
 
   The most obvious reason being that it's easier for the translator to
   decide how to translate it.
   
   The more subtle but non the less much more important one is: 
       xgettext, the actual message extraction tool, creates only one
       translatable catalog entry for identical strings. Thus you'd get
       only one "s", where the translator would need different entries
       because the translations don't end the same.
   
   

That's all for now. I think at least part of these points should go into a
file in the toplevel dir, something like README.Coders :)


 
Philipp

-- 
You have moved your mouse. Windows must be rebooted for the
changes to take effect.

Re: Notes on i18n

[Jeffrey A Law]

  In message < 37438716.244329@mailer.gwdg.de >you write:
  > As the 2.95 branch has happened, it's time to unearth ones todo lists. And
  > here is part of mine (other parts will follow in separate posts).
  > 
  > Before I start, a warning for the casual reader: this *will* get rather
  > long, so if you're not actively taking part in egcs's development,
  > you may safely skip this message.
Note that our focus will still be on gcc-2.95 until it is released.  I would
much prefer everyone to keep that in mind and work towards the release rather
than working on new stuff.

Remember, the release will only be as good as the folks on the list make it.
The more people working on testing and debugging of the release over the next
6 weeks the better the release will be.

jeff

Re: Notes on i18n

[Dean Wilson]

On Wed, 19 May 1999, Jeffrey A Law wrote:

> 
> 
>   In message < 37438716.244329@mailer.gwdg.de >you write:
>   > As the 2.95 branch has happened, it's time to unearth ones todo lists. And
>   > here is part of mine (other parts will follow in separate posts).
>   > 
>   > Before I start, a warning for the casual reader: this *will* get rather
>   > long, so if you're not actively taking part in egcs's development,
>   > you may safely skip this message.
> Note that our focus will still be on gcc-2.95 until it is released.  I would
> much prefer everyone to keep that in mind and work towards the release rather
> than working on new stuff.
> 
> Remember, the release will only be as good as the folks on the list make it.
> The more people working on testing and debugging of the release over the next
> 6 weeks the better the release will be.
> 
> jeff
> 

volunteer here still, very board hoping to change that. im yours to
command, so to speak, with lots of time to kill as it happens till mid
july. 

Dean Wilson

Re: Regression testing

[Jeffrey A Law]

  In message < Pine.LNX.3.96.990519223526.10106C-100000@speedbros.com >you write:
  > volunteer here still, very board hoping to change that. im yours to
  > command, so to speak, with lots of time to kill as it happens till mid
  > july. 
Chee-Wai Yeung has volunteered to help coordinate the regression testing
effort.  I've still got to do the initial organization (and it is safe to
assume that I am quite busy.  This takes time to set up).

You can get an idea of what you need to do by reading:

http://egcs.cygnus.com/gcc-2.95/regress.html

You will need to inform Chee-Wai Yeung what platforms off the test matrix
on that page that you will volunteer to test.

The most important thing is to get test results for egcs-1.1.x based compilers
first since those results are not going to change.

Also note that we want to regression test all our compilers & runtime libraries
that have testsuites.  ie, C, C++, F77, ObjC, & Java with their associated
libraries (libstdc++, libio, libf2c, libobjc & libjava).

jeff

Re: Regression test volunteers for gcc-2.95

[Jeffrey A Law]

It's time to ask for volunteers to perform regression testing of the
gcc-2.95 release.

Before you volunteer, I suggest you read

http://egcs.cygnus.com/gcc-2.95/regress.html

Note carefully that we need folks that can test *both* an egcs-1.1 based
compiler as well as a gcc-2.95 based compiler.  The basic idea is for
platforms noted on that page we should not have any regressions when we
compare the results of the egcs-1.1.x compiler to the results of running
the same testsuite with a gcc-2.95 compiler.

The most important thing to do now is get the list of volunteers and for
the volunteers to get the egcs-1.1.x based results which provide the baseline
for comparisons with gcc-2.95.

I will go ahead and volunteer to regression test hppa1.1-hp-hpux10.20 and
m68k-hp-bsd4.4.

Cheewai Yeung cheewai_yeung@yahoo.com will be coordinating the analysis
of the test results.  You should send him mail if you can volunteer to test
one of the listed platforms.

Questions about the process should be sent to both Cheewai and myself (mostly
so that I can try to clarify on the web pages any issues with the procedures
for performing the tests and reporting the results).

jeff

Re: Regression test volunteers for gcc-2.95

[Joe Buck]

Jeff writes:
> The most important thing to do now is get the list of volunteers and for
> the volunteers to get the egcs-1.1.x based results which provide the baseline
> for comparisons with gcc-2.95.

I strongly suggest that you require egcs-1.1.2, not egcs-1.1.x, as the
reference release for regressions.  Otherwise we will be comparing apples
and oranges.

Joe

Re: Regression test volunteers for gcc-2.95

[David Starner]

Jeffrey A Law wrote:
> 
> It's time to ask for volunteers to perform regression testing of the
> gcc-2.95 release.

Is there any point in testing sparc-sun-solaris2.5.1? If there is, would
regression testing be helpful, or would just testing 2.95 be acceptable?

Re: Regression test volunteers for gcc-2.95

[Joe Buck]

> Is there any point in testing sparc-sun-solaris2.5.1? If there is, would
> regression testing be helpful, or would just testing 2.95 be acceptable?

Certainly; Solaris 2.5.1 is still in wide use (e.g. I'm using it now,
and I'm wider than I was when I was younger ;-).

For past releases, I submitted tests for that platform, and I can still
do it.

Re: Regression test volunteers for gcc-2.95

[Jeffrey A Law]

  In message < 199905201654.JAA06245@atrus.synopsys.com >you write:
  > I strongly suggest that you require egcs-1.1.2, not egcs-1.1.x, as the
  > reference release for regressions.  Otherwise we will be comparing apples
  > and oranges.
Yes.  That makes perfect sense.  I'll update the web pages as soon as
I can (I'm in transit to LinuxExpo at the moment).

Jeff

Re: Regression test volunteers for gcc-2.95

[Jerry Quinn]

Jeffrey A Law wrote:
> 
> It's time to ask for volunteers to perform regression testing of the
> gcc-2.95 release.
 
> I will go ahead and volunteer to regression test hppa1.1-hp-hpux10.20 and
> m68k-hp-bsd4.4.

Do we need regression on hppa2.0-hp-hpux10.20?  1.1.2 didn't have a
distinct 2.0 configuration, so it's effectively a new platform, although
for the most part it's still identical (assuming we don't build with
-march=2.0).


-- 
Jerry Quinn                             Tel: (514) 761-8737
jquinn@nortelnetworks.com               Fax: (514) 761-8505
Speech Recognition Research

Re: Regression test volunteers for gcc-2.95

[Jeffrey A Law]

  In message < 199905202044.NAA14661@atrus.synopsys.com >you write:
  > 
  > > Is there any point in testing sparc-sun-solaris2.5.1? If there is, would
  > > regression testing be helpful, or would just testing 2.95 be acceptable?
  > 
  > Certainly; Solaris 2.5.1 is still in wide use (e.g. I'm using it now,
  > and I'm wider than I was when I was younger ;-).
  > 
  > For past releases, I submitted tests for that platform, and I can still
  > do it.
Note we can't put everything on the matrix and guarantee that we'll fix
regressions.  I debated 2.5/2.6 2.5/2.7 and 2.6/2.7 and decided 2.6/2.7 was
the best for no particular reason other than we need 2.7 tested and 2.6 is
quite popular.  Testing 3 solaris platforms seemed like overkill :-)


We still want to fix any bugs that are exposed, even on platforms not on the
regression test matrix.  But the platforms on the matrix will have priority
over those not on the matrix.

So feel free to send in test results for platforms not on the regression
matrix.  Just be aware of what platforms get top priority :-)

jeff

Re: Regression test volunteers for gcc-2.95

[Jeffrey A Law]

  In message < 374569F9.88E1E822@americasm01.nt.com >you write:
  > Jeffrey A Law wrote:
  > > 
  > > It's time to ask for volunteers to perform regression testing of the
  > > gcc-2.95 release.
  >  
  > > I will go ahead and volunteer to regression test hppa1.1-hp-hpux10.20 and
  > > m68k-hp-bsd4.4.
  > 
  > Do we need regression on hppa2.0-hp-hpux10.20?  1.1.2 didn't have a
  > distinct 2.0 configuration, so it's effectively a new platform, although
  > for the most part it's still identical (assuming we don't build with
  > -march=2.0).
Good question.  I think the best thing to do is to compare both
hppa2.0-hp-hpux10.20 and hppa1.1-hp-hpux10.20 for gcc-2.95 against
hppa1.1-hp-hpux10.20 for egcs-1.1.2.

I've still got access to not too terribly slow hppa1.1 machines, so I'll
provide results for the hppa1.1 platforms.  I suspect the gcc-2.95 series
will be the last ones where we regression test for hppa1.1.

jeff

Re: Regression test volunteers for gcc-2.95

[Jeffrey A Law]

  In message < 374569F9.88E1E822@americasm01.nt.com >you write:
  > Jeffrey A Law wrote:
  > > 
  > > It's time to ask for volunteers to perform regression testing of the
  > > gcc-2.95 release.
  >  
  > > I will go ahead and volunteer to regression test hppa1.1-hp-hpux10.20 and
  > > m68k-hp-bsd4.4.
  > 
  > Do we need regression on hppa2.0-hp-hpux10.20?  1.1.2 didn't have a
  > distinct 2.0 configuration, so it's effectively a new platform, although
  > for the most part it's still identical (assuming we don't build with
  > -march=2.0).
Good question.  I think the best thing to do is to compare both
hppa2.0-hp-hpux10.20 and hppa1.1-hp-hpux10.20 for gcc-2.95 against
hppa1.1-hp-hpux10.20 for egcs-1.1.2.

I've still got access to not too terribly slow hppa1.1 machines, so I'll
provide results for the hppa1.1 platforms.  I suspect the gcc-2.95 series
will be the last ones where we regression test for hppa1.1.

jeff

Re: Regression test volunteers for gcc-2.95

[Jeffrey A Law]

  In message < 199905202044.NAA14661@atrus.synopsys.com >you write:
  > 
  > > Is there any point in testing sparc-sun-solaris2.5.1? If there is, would
  > > regression testing be helpful, or would just testing 2.95 be acceptable?
  > 
  > Certainly; Solaris 2.5.1 is still in wide use (e.g. I'm using it now,
  > and I'm wider than I was when I was younger ;-).
  > 
  > For past releases, I submitted tests for that platform, and I can still
  > do it.
Note we can't put everything on the matrix and guarantee that we'll fix
regressions.  I debated 2.5/2.6 2.5/2.7 and 2.6/2.7 and decided 2.6/2.7 was
the best for no particular reason other than we need 2.7 tested and 2.6 is
quite popular.  Testing 3 solaris platforms seemed like overkill :-)


We still want to fix any bugs that are exposed, even on platforms not on the
regression test matrix.  But the platforms on the matrix will have priority
over those not on the matrix.

So feel free to send in test results for platforms not on the regression
matrix.  Just be aware of what platforms get top priority :-)

jeff

Re: Regression test volunteers for gcc-2.95

[Joe Buck]

> Is there any point in testing sparc-sun-solaris2.5.1? If there is, would
> regression testing be helpful, or would just testing 2.95 be acceptable?

Certainly; Solaris 2.5.1 is still in wide use (e.g. I'm using it now,
and I'm wider than I was when I was younger ;-).

For past releases, I submitted tests for that platform, and I can still
do it.

Notes on i18n

[Philipp Thomas]

As the 2.95 branch has happened, it's time to unearth ones todo lists. And
here is part of mine (other parts will follow in separate posts).

Before I start, a warning for the casual reader: this *will* get rather
long, so if you're not actively taking part in egcs's development,
you may safely skip this message.

	
While working on completing i18n for egcs I've made some observations.
Some of them are not always obvious for those whose native tongue is English,
others I'll mention just to make it simpler for those that haven't dealt
with i18n matters before.

I'd like to share them all with you fellow egcs developers in order to ease my
work (well, wouldn't you have guessed :) and that of other future translators.

In the following I'll abbreviate gcc/ABOUT-GCC-NLS as AGN.

 - Before you implement any diagnostic/warning/error reporting functions,
   please read AGN for a description of the mechanisms used, and for
   guidelines as to what should and what shouldn't be translated.

 - I'd also recommend browsing gettext.texi - part of the GNU gettext
   package - specifically the topics under `Preparing Program Sources' in
   the toplevel menu.

 - if you create a new file, or change an existing one that contains/will
   contain something that should be translated, please update
   gcc/po/POTFILES.in if necessary.

 - In functions that are part of the reporting mechanism of the
   frontend (f.i. cp_error for g++), please name the parameter in which the
   message string is passed msgid or end its name with msgid. gcc/exgettext,
   gcc's helper script searches for these names to determine which strings
   to extract for translation.
   
 - When you create/use static message strings, encapsulate them either with
   N_() if the string will get translated on a higher level, or with _()
   if the string is used locally to construct a larger message.
   This makes the strings visible for the message extraction tool(s).
   Here an example from java/parse.c (out of my to_send queue :) :

      char *msg = (!valid_cast_to_p (type_value, type) ?
		   _("Can't") : _("Explicit cast needed to"));

 - If you must use printf et al, mark the message with _(). Again, please see
   to it that direct printfs are concentrated in as few routines as possible
   in order to avoid cluttering up the sources with _().

 - When using conditional expressions for printing the plural form, _please_
   use the full words, rather than just the endings, e.g. 
   
     printf ("%d %s\n", n, (n == 1 ? "warning" : "warnings");
   
   instead of 
   
     printf("%d warning%s\n", n, (n == 1 ? "" : "s")
 
   The most obvious reason being that it's easier for the translator to
   decide how to translate it.
   
   The more subtle but non the less much more important one is: 
       xgettext, the actual message extraction tool, creates only one
       translatable catalog entry for identical strings. Thus you'd get
       only one "s", where the translator would need different entries
       because the translations don't end the same.
   
   

That's all for now. I think at least part of these points should go into a
file in the toplevel dir, something like README.Coders :)


 
Philipp

-- 
You have moved your mouse. Windows must be rebooted for the
changes to take effect.

Re: Regression test volunteers for gcc-2.95

[Jeffrey A Law]

  In message < 199905201654.JAA06245@atrus.synopsys.com >you write:
  > I strongly suggest that you require egcs-1.1.2, not egcs-1.1.x, as the
  > reference release for regressions.  Otherwise we will be comparing apples
  > and oranges.
Yes.  That makes perfect sense.  I'll update the web pages as soon as
I can (I'm in transit to LinuxExpo at the moment).

Jeff

Re: Regression testing

[Jeffrey A Law]

  In message < Pine.LNX.3.96.990519223526.10106C-100000@speedbros.com >you write:
  > volunteer here still, very board hoping to change that. im yours to
  > command, so to speak, with lots of time to kill as it happens till mid
  > july. 
Chee-Wai Yeung has volunteered to help coordinate the regression testing
effort.  I've still got to do the initial organization (and it is safe to
assume that I am quite busy.  This takes time to set up).

You can get an idea of what you need to do by reading:

http://egcs.cygnus.com/gcc-2.95/regress.html

You will need to inform Chee-Wai Yeung what platforms off the test matrix
on that page that you will volunteer to test.

The most important thing is to get test results for egcs-1.1.x based compilers
first since those results are not going to change.

Also note that we want to regression test all our compilers & runtime libraries
that have testsuites.  ie, C, C++, F77, ObjC, & Java with their associated
libraries (libstdc++, libio, libf2c, libobjc & libjava).

jeff

Re: Notes on i18n

[Dean Wilson]

On Wed, 19 May 1999, Jeffrey A Law wrote:

> 
> 
>   In message < 37438716.244329@mailer.gwdg.de >you write:
>   > As the 2.95 branch has happened, it's time to unearth ones todo lists. And
>   > here is part of mine (other parts will follow in separate posts).
>   > 
>   > Before I start, a warning for the casual reader: this *will* get rather
>   > long, so if you're not actively taking part in egcs's development,
>   > you may safely skip this message.
> Note that our focus will still be on gcc-2.95 until it is released.  I would
> much prefer everyone to keep that in mind and work towards the release rather
> than working on new stuff.
> 
> Remember, the release will only be as good as the folks on the list make it.
> The more people working on testing and debugging of the release over the next
> 6 weeks the better the release will be.
> 
> jeff
> 

volunteer here still, very board hoping to change that. im yours to
command, so to speak, with lots of time to kill as it happens till mid
july. 

Dean Wilson

Re: Regression test volunteers for gcc-2.95

[Jerry Quinn]

Jeffrey A Law wrote:
> 
> It's time to ask for volunteers to perform regression testing of the
> gcc-2.95 release.
 
> I will go ahead and volunteer to regression test hppa1.1-hp-hpux10.20 and
> m68k-hp-bsd4.4.

Do we need regression on hppa2.0-hp-hpux10.20?  1.1.2 didn't have a
distinct 2.0 configuration, so it's effectively a new platform, although
for the most part it's still identical (assuming we don't build with
-march=2.0).


-- 
Jerry Quinn                             Tel: (514) 761-8737
jquinn@nortelnetworks.com               Fax: (514) 761-8505
Speech Recognition Research

Re: Regression test volunteers for gcc-2.95

[Joe Buck]

Jeff writes:
> The most important thing to do now is get the list of volunteers and for
> the volunteers to get the egcs-1.1.x based results which provide the baseline
> for comparisons with gcc-2.95.

I strongly suggest that you require egcs-1.1.2, not egcs-1.1.x, as the
reference release for regressions.  Otherwise we will be comparing apples
and oranges.

Joe

Re: Notes on i18n

[Jeffrey A Law]

  In message < 37438716.244329@mailer.gwdg.de >you write:
  > As the 2.95 branch has happened, it's time to unearth ones todo lists. And
  > here is part of mine (other parts will follow in separate posts).
  > 
  > Before I start, a warning for the casual reader: this *will* get rather
  > long, so if you're not actively taking part in egcs's development,
  > you may safely skip this message.
Note that our focus will still be on gcc-2.95 until it is released.  I would
much prefer everyone to keep that in mind and work towards the release rather
than working on new stuff.

Remember, the release will only be as good as the folks on the list make it.
The more people working on testing and debugging of the release over the next
6 weeks the better the release will be.

jeff

Re: Regression test volunteers for gcc-2.95

[David Starner]

Jeffrey A Law wrote:
> 
> It's time to ask for volunteers to perform regression testing of the
> gcc-2.95 release.

Is there any point in testing sparc-sun-solaris2.5.1? If there is, would
regression testing be helpful, or would just testing 2.95 be acceptable?

Re: Regression test volunteers for gcc-2.95

[Jeffrey A Law]

It's time to ask for volunteers to perform regression testing of the
gcc-2.95 release.

Before you volunteer, I suggest you read

http://egcs.cygnus.com/gcc-2.95/regress.html

Note carefully that we need folks that can test *both* an egcs-1.1 based
compiler as well as a gcc-2.95 based compiler.  The basic idea is for
platforms noted on that page we should not have any regressions when we
compare the results of the egcs-1.1.x compiler to the results of running
the same testsuite with a gcc-2.95 compiler.

The most important thing to do now is get the list of volunteers and for
the volunteers to get the egcs-1.1.x based results which provide the baseline
for comparisons with gcc-2.95.

I will go ahead and volunteer to regression test hppa1.1-hp-hpux10.20 and
m68k-hp-bsd4.4.

Cheewai Yeung cheewai_yeung@yahoo.com will be coordinating the analysis
of the test results.  You should send him mail if you can volunteer to test
one of the listed platforms.

Questions about the process should be sent to both Cheewai and myself (mostly
so that I can try to clarify on the web pages any issues with the procedures
for performing the tests and reporting the results).

jeff

Re: Regression test volunteers for gcc-2.95

[Chee-Wai Yeung]

Jerry

I would say go ahead and also regress this one.


Jeff, any concerns with this?

Regards,
Chee Wai

--- Jerry Quinn <jquinn@nortelnetworks.com> wrote:
> Jeffrey A Law wrote:
> > 
> > It's time to ask for volunteers to perform
> regression testing of the
> > gcc-2.95 release.
>  
> > I will go ahead and volunteer to regression test
> hppa1.1-hp-hpux10.20 and
> > m68k-hp-bsd4.4.
> 
> Do we need regression on hppa2.0-hp-hpux10.20? 
> 1.1.2 didn't have a
> distinct 2.0 configuration, so it's effectively a
> new platform, although
> for the most part it's still identical (assuming we
> don't build with
> -march=2.0).
> 
> 
> -- 
> Jerry Quinn                             Tel: (514)
> 761-8737
> jquinn@nortelnetworks.com               Fax: (514)
> 761-8505
> Speech Recognition Research
> 

===
~~~~~~~~~~~~~~
Chee-Wai Yeung
Email: cheewai_yeung@yahoo.com
       cwyeung@hknet.com

_____________________________________________________________
Do You Yahoo!?
Free instant messaging and more at http://messenger.yahoo.com

Re: Regression test volunteers for gcc-2.95

[Chee-Wai Yeung]

Jerry

I would say go ahead and also regress this one.


Jeff, any concerns with this?

Regards,
Chee Wai

--- Jerry Quinn <jquinn@nortelnetworks.com> wrote:
> Jeffrey A Law wrote:
> > 
> > It's time to ask for volunteers to perform
> regression testing of the
> > gcc-2.95 release.
>  
> > I will go ahead and volunteer to regression test
> hppa1.1-hp-hpux10.20 and
> > m68k-hp-bsd4.4.
> 
> Do we need regression on hppa2.0-hp-hpux10.20? 
> 1.1.2 didn't have a
> distinct 2.0 configuration, so it's effectively a
> new platform, although
> for the most part it's still identical (assuming we
> don't build with
> -march=2.0).
> 
> 
> -- 
> Jerry Quinn                             Tel: (514)
> 761-8737
> jquinn@nortelnetworks.com               Fax: (514)
> 761-8505
> Speech Recognition Research
> 

===
~~~~~~~~~~~~~~
Chee-Wai Yeung
Email: cheewai_yeung@yahoo.com
       cwyeung@hknet.com

_____________________________________________________________
Do You Yahoo!?
Free instant messaging and more at http://messenger.yahoo.com