[Cluster-devel] [PATCH 3/3] mkfs i18n:

[Cluster-devel] [PATCH 3/3] mkfs i18n:

[Carlos Maiolino]

Use rpmatch() function to get answer to
yes-or-no questions during mkfs confirmation

Signed-off-by: Carlos Maiolino <cmaiolino@redhat.com>
---
 gfs2/mkfs/main_mkfs.c |   30 ++++++++++++++++++++++--------
 1 files changed, 22 insertions(+), 8 deletions(-)

diff --git a/gfs2/mkfs/main_mkfs.c b/gfs2/mkfs/main_mkfs.c
index 4751f19..d68cf98 100644
--- a/gfs2/mkfs/main_mkfs.c
+++ b/gfs2/mkfs/main_mkfs.c
@@ -456,8 +456,11 @@ fail:
 
 static void are_you_sure(struct gfs2_sbd *sdp)
 {
-	char input[32];
+	char *line = NULL;
+	size_t len = 0;
 	int fd;
+	int ret = -1;
+	int res = 0;
 
 	fd = open(sdp->device_name, O_RDONLY|O_CLOEXEC);
 	if (fd < 0)
@@ -465,14 +468,25 @@ static void are_you_sure(struct gfs2_sbd *sdp)
 	printf( _("This will destroy any data on %s.\n"), sdp->device_name);
 	check_dev_content(sdp->device_name);
 	close(fd);
-	printf( _("\nAre you sure you want to proceed? [y/n] "));
-	if(!fgets(input, 32, stdin))
-		die( _("unable to read from stdin\n"));
+	
+	do{
+		printf( _("\nAre you sure you want to proceed? [y/n]"));
+		ret = getline(&line, &len, stdin);
+		res = rpmatch(line);
+		
+		if (res > 0){
+			free(line);
+			return;
+		}
+		if (!res){
+			printf("\n");
+			die( _("aborted\n"));
+		}
+		
+	}while(ret >= 0);
 
-	if (input[0] != 'y')
-		die( _("aborted\n"));
-	else
-		printf("\n");
+	if(line)
+		free(line);
 }
 
 /**
-- 
1.7.6

[Cluster-devel] [PATCH 3/3] mkfs i18n:

[Bob Peterson]

Hi Carlos,

----- Original Message -----
| + do{
| + printf( _("\nAre you sure you want to proceed? [y/n]"));
| + ret = getline(&line, &len, stdin);
| + res = rpmatch(line);

I'd like to see the free(line); moved right here after rpmatch.
Otherwise multiple iterations through the loop will leak memory.

| +
| + if (res > 0){

With my suggestion above, this can be removed:
| + free(line);

Other than that, it looks good.

Regards,

Bob Peterson
Red Hat File Systems

[Cluster-devel] [PATCH 3/3] mkfs i18n:

[Steven Whitehouse]

Hi,

On Tue, 2011-09-06 at 14:41 -0300, Carlos Maiolino wrote:
> Use rpmatch() function to get answer to
> yes-or-no questions during mkfs confirmation
> 
> Signed-off-by: Carlos Maiolino <cmaiolino@redhat.com>
> ---
>  gfs2/mkfs/main_mkfs.c |   30 ++++++++++++++++++++++--------
>  1 files changed, 22 insertions(+), 8 deletions(-)
> 
> diff --git a/gfs2/mkfs/main_mkfs.c b/gfs2/mkfs/main_mkfs.c
> index 4751f19..d68cf98 100644
> --- a/gfs2/mkfs/main_mkfs.c
> +++ b/gfs2/mkfs/main_mkfs.c
> @@ -456,8 +456,11 @@ fail:
>  
>  static void are_you_sure(struct gfs2_sbd *sdp)
>  {
> -	char input[32];
> +	char *line = NULL;
> +	size_t len = 0;
>  	int fd;
> +	int ret = -1;
> +	int res = 0;
>  
>  	fd = open(sdp->device_name, O_RDONLY|O_CLOEXEC);
>  	if (fd < 0)
> @@ -465,14 +468,25 @@ static void are_you_sure(struct gfs2_sbd *sdp)
>  	printf( _("This will destroy any data on %s.\n"), sdp->device_name);
>  	check_dev_content(sdp->device_name);
>  	close(fd);
> -	printf( _("\nAre you sure you want to proceed? [y/n] "));
> -	if(!fgets(input, 32, stdin))
> -		die( _("unable to read from stdin\n"));
> +	
> +	do{
> +		printf( _("\nAre you sure you want to proceed? [y/n]"));
This means that the translator needs to know that rpmatch is being used
and to also know how to translate this accordingly. Why not use
nl_langinfo() to put the response strings into this message?

Also, starting the string with \n is a bit odd too,

Steve.


> +		ret = getline(&line, &len, stdin);
> +		res = rpmatch(line);
> +		
> +		if (res > 0){
> +			free(line);
> +			return;
> +		}
> +		if (!res){
> +			printf("\n");
> +			die( _("aborted\n"));
> +		}
> +		
> +	}while(ret >= 0);
>  
> -	if (input[0] != 'y')
> -		die( _("aborted\n"));
> -	else
> -		printf("\n");
> +	if(line)
> +		free(line);
>  }
>  
>  /**

[Cluster-devel] [PATCH 3/3] mkfs i18n:

[Carlos Maiolino]

Hi,
On Tue, Sep 06, 2011 at 08:10:06PM +0100, Steven Whitehouse wrote:
> Hi,
> 
> On Tue, 2011-09-06 at 14:41 -0300, Carlos Maiolino wrote:
> > Use rpmatch() function to get answer to
> > yes-or-no questions during mkfs confirmation
> > 
> > Signed-off-by: Carlos Maiolino <cmaiolino@redhat.com>
> > ---
> >  gfs2/mkfs/main_mkfs.c |   30 ++++++++++++++++++++++--------
> >  1 files changed, 22 insertions(+), 8 deletions(-)
> > 
> > diff --git a/gfs2/mkfs/main_mkfs.c b/gfs2/mkfs/main_mkfs.c
> > index 4751f19..d68cf98 100644
> > --- a/gfs2/mkfs/main_mkfs.c
> > +++ b/gfs2/mkfs/main_mkfs.c
> > @@ -456,8 +456,11 @@ fail:
> >  
> >  static void are_you_sure(struct gfs2_sbd *sdp)
> >  {
> > -	char input[32];
> > +	char *line = NULL;
> > +	size_t len = 0;
> >  	int fd;
> > +	int ret = -1;
> > +	int res = 0;
> >  
> >  	fd = open(sdp->device_name, O_RDONLY|O_CLOEXEC);
> >  	if (fd < 0)
> > @@ -465,14 +468,25 @@ static void are_you_sure(struct gfs2_sbd *sdp)
> >  	printf( _("This will destroy any data on %s.\n"), sdp->device_name);
> >  	check_dev_content(sdp->device_name);
> >  	close(fd);
> > -	printf( _("\nAre you sure you want to proceed? [y/n] "));
> > -	if(!fgets(input, 32, stdin))
> > -		die( _("unable to read from stdin\n"));
> > +	
> > +	do{
> > +		printf( _("\nAre you sure you want to proceed? [y/n]"));
> This means that the translator needs to know that rpmatch is being used
> and to also know how to translate this accordingly. Why not use
> nl_langinfo() to put the response strings into this message?
> 
I was looking how nl_langinfo() works, and all it does to help with yes/no 
questions is to return a regex expression (like [^[SsyY].*) to be used to match 
the user's answer, it does not return a string or char value like yes/no (en_US) or 
sim/nao(pt_BR) we can use to replace the y/n in the question string.

The rpmatch() function takes advantage of nl_langinfo() internaly to match the
customer's answer according to which l10n the system is set.

This way, even using nl_langinfo(), the translator person will still need to translate
the y/n to s/n, since is not possible to use nl_langinfo to replace these values.

What we could do, is to create an auxiliar function to check the return of nl_langinfo(),
and, according with this value, write an specific string (s/n, y/n, etc). But, imho, this
could lead to an error prone condition during string creation.

What u think?

Cheers,
-- 
--Carlos

[Cluster-devel] [PATCH 3/3] mkfs i18n:

[Carlos Maiolino]

Hi Steve, I was looking through some codes from GNU project, and here are some examples I found:


from fdisk tool (this is the latest version, but I think fdisk is not often updated):

/* Prompt the user to answer a yes or no question. */
int
command_line_prompt_boolean_question (const char* prompt) {
        char        *user_ans;
        StrList *possibilities = str_list_create (_("yes"), _("no"), NULL);
        user_ans = fdisk_command_line_get_word (_(prompt), _("no"), possibilities, 0);

        if (strcmp (user_ans, _("yes")) == 0)
                return 1;

        /* user answered no */
        return 0;
}


I found another idea in the gnu parted.

The GNU parted adds the rpmatch function as a library
into the parted source code, and adds a regex expression
to be translated into the .pot file:

<snip>
 #: lib/rpmatch.c:147
 msgid "^[yY]"
 msgstr ""

 #: lib/rpmatch.c:160
 msgid "^[nN]"
 msgstr ""

</snip>

Looking through both of the above codes, looks like there 
is no specific rule about how to translate "y/n" strings.
In both cases these are translated via .pot file.

One suggestion, would be to split the string

"Are you sure? [y/n]"

in three different strings like:

"Are you sure?"
"yes"
"no"

This way would be more 'understandable' for translators
than the current one.
Or, we can add to the .pot file a comment explaining 
what to do with this portion of translation.



> > > Signed-off-by: Carlos Maiolino <cmaiolino@redhat.com>
> > > ---
> > >  gfs2/mkfs/main_mkfs.c |   30 ++++++++++++++++++++++--------
> > >  1 files changed, 22 insertions(+), 8 deletions(-)
> > > 
> > > diff --git a/gfs2/mkfs/main_mkfs.c b/gfs2/mkfs/main_mkfs.c
> > > index 4751f19..d68cf98 100644
> > > --- a/gfs2/mkfs/main_mkfs.c
> > > +++ b/gfs2/mkfs/main_mkfs.c
> > > @@ -456,8 +456,11 @@ fail:
> > >  
> > >  static void are_you_sure(struct gfs2_sbd *sdp)
> > >  {
> > > -	char input[32];
> > > +	char *line = NULL;
> > > +	size_t len = 0;
> > >  	int fd;
> > > +	int ret = -1;
> > > +	int res = 0;
> > >  
> > >  	fd = open(sdp->device_name, O_RDONLY|O_CLOEXEC);
> > >  	if (fd < 0)
> > > @@ -465,14 +468,25 @@ static void are_you_sure(struct gfs2_sbd *sdp)
> > >  	printf( _("This will destroy any data on %s.\n"), sdp->device_name);
> > >  	check_dev_content(sdp->device_name);
> > >  	close(fd);
> > > -	printf( _("\nAre you sure you want to proceed? [y/n] "));
> > > -	if(!fgets(input, 32, stdin))
> > > -		die( _("unable to read from stdin\n"));
> > > +	
> > > +	do{
> > > +		printf( _("\nAre you sure you want to proceed? [y/n]"));
> > This means that the translator needs to know that rpmatch is being used
> > and to also know how to translate this accordingly. Why not use
> > nl_langinfo() to put the response strings into this message?
> > 
> I was looking how nl_langinfo() works, and all it does to help with yes/no 
> questions is to return a regex expression (like [^[SsyY].*) to be used to match 
> the user's answer, it does not return a string or char value like yes/no (en_US) or 
> sim/nao(pt_BR) we can use to replace the y/n in the question string.
> 
> The rpmatch() function takes advantage of nl_langinfo() internaly to match the
> customer's answer according to which l10n the system is set.
> 
> This way, even using nl_langinfo(), the translator person will still need to translate
> the y/n to s/n, since is not possible to use nl_langinfo to replace these values.
> 
> What we could do, is to create an auxiliar function to check the return of nl_langinfo(),
> and, according with this value, write an specific string (s/n, y/n, etc). But, imho, this
> could lead to an error prone condition during string creation.
> 
> What u think?
> 
> Cheers,
> -- 
> --Carlos
> 

-- 
--Carlos

[Cluster-devel] [PATCH 3/3] mkfs i18n:

[Steven Whitehouse]

Hi,

On Tue, 2011-09-20 at 21:33 -0300, Carlos Maiolino wrote:
> Hi Steve, I was looking through some codes from GNU project, and here are some examples I found:
> 
> 
> from fdisk tool (this is the latest version, but I think fdisk is not often updated):
> 
> /* Prompt the user to answer a yes or no question. */
> int
> command_line_prompt_boolean_question (const char* prompt) {
>         char        *user_ans;
>         StrList *possibilities = str_list_create (_("yes"), _("no"), NULL);
>         user_ans = fdisk_command_line_get_word (_(prompt), _("no"), possibilities, 0);
> 
>         if (strcmp (user_ans, _("yes")) == 0)
>                 return 1;
> 
>         /* user answered no */
>         return 0;
> }
> 
> 
> I found another idea in the gnu parted.
> 
> The GNU parted adds the rpmatch function as a library
> into the parted source code, and adds a regex expression
> to be translated into the .pot file:
> 
> <snip>
>  #: lib/rpmatch.c:147
>  msgid "^[yY]"
>  msgstr ""
> 
>  #: lib/rpmatch.c:160
>  msgid "^[nN]"
>  msgstr ""
> 
> </snip>
> 
> Looking through both of the above codes, looks like there 
> is no specific rule about how to translate "y/n" strings.
> In both cases these are translated via .pot file.
> 
> One suggestion, would be to split the string
> 
> "Are you sure? [y/n]"
> 
> in three different strings like:
> 
> "Are you sure?"
> "yes"
> "no"
> 
> This way would be more 'understandable' for translators
> than the current one.
> Or, we can add to the .pot file a comment explaining 
> what to do with this portion of translation.
> 
> 
I prefer the longer string to splitting it, if we cannot get the
translation automatically from glibc, which does appear to be the case.
Also, I think adding a comment to the .pot to explain how to translate
the question is a good idea. We should always be doing that where
anything tricky is encountered,

Steve.

> 
> > > > Signed-off-by: Carlos Maiolino <cmaiolino@redhat.com>
> > > > ---
> > > >  gfs2/mkfs/main_mkfs.c |   30 ++++++++++++++++++++++--------
> > > >  1 files changed, 22 insertions(+), 8 deletions(-)
> > > > 
> > > > diff --git a/gfs2/mkfs/main_mkfs.c b/gfs2/mkfs/main_mkfs.c
> > > > index 4751f19..d68cf98 100644
> > > > --- a/gfs2/mkfs/main_mkfs.c
> > > > +++ b/gfs2/mkfs/main_mkfs.c
> > > > @@ -456,8 +456,11 @@ fail:
> > > >  
> > > >  static void are_you_sure(struct gfs2_sbd *sdp)
> > > >  {
> > > > -	char input[32];
> > > > +	char *line = NULL;
> > > > +	size_t len = 0;
> > > >  	int fd;
> > > > +	int ret = -1;
> > > > +	int res = 0;
> > > >  
> > > >  	fd = open(sdp->device_name, O_RDONLY|O_CLOEXEC);
> > > >  	if (fd < 0)
> > > > @@ -465,14 +468,25 @@ static void are_you_sure(struct gfs2_sbd *sdp)
> > > >  	printf( _("This will destroy any data on %s.\n"), sdp->device_name);
> > > >  	check_dev_content(sdp->device_name);
> > > >  	close(fd);
> > > > -	printf( _("\nAre you sure you want to proceed? [y/n] "));
> > > > -	if(!fgets(input, 32, stdin))
> > > > -		die( _("unable to read from stdin\n"));
> > > > +	
> > > > +	do{
> > > > +		printf( _("\nAre you sure you want to proceed? [y/n]"));
> > > This means that the translator needs to know that rpmatch is being used
> > > and to also know how to translate this accordingly. Why not use
> > > nl_langinfo() to put the response strings into this message?
> > > 
> > I was looking how nl_langinfo() works, and all it does to help with yes/no 
> > questions is to return a regex expression (like [^[SsyY].*) to be used to match 
> > the user's answer, it does not return a string or char value like yes/no (en_US) or 
> > sim/nao(pt_BR) we can use to replace the y/n in the question string.
> > 
> > The rpmatch() function takes advantage of nl_langinfo() internaly to match the
> > customer's answer according to which l10n the system is set.
> > 
> > This way, even using nl_langinfo(), the translator person will still need to translate
> > the y/n to s/n, since is not possible to use nl_langinfo to replace these values.
> > 
> > What we could do, is to create an auxiliar function to check the return of nl_langinfo(),
> > and, according with this value, write an specific string (s/n, y/n, etc). But, imho, this
> > could lead to an error prone condition during string creation.
> > 
> > What u think?
> > 
> > Cheers,
> > -- 
> > --Carlos
> > 
> 

[Cluster-devel] [PATCH 3/3] mkfs i18n:

[Steven Whitehouse]

Hi,

On Tue, 2011-09-20 at 14:27 -0300, Carlos Maiolino wrote:
> Hi,
> On Tue, Sep 06, 2011 at 08:10:06PM +0100, Steven Whitehouse wrote:
> > Hi,
> > 
> > On Tue, 2011-09-06 at 14:41 -0300, Carlos Maiolino wrote:
> > > Use rpmatch() function to get answer to
> > > yes-or-no questions during mkfs confirmation
> > > 
> > > Signed-off-by: Carlos Maiolino <cmaiolino@redhat.com>
> > > ---
> > >  gfs2/mkfs/main_mkfs.c |   30 ++++++++++++++++++++++--------
> > >  1 files changed, 22 insertions(+), 8 deletions(-)
> > > 
> > > diff --git a/gfs2/mkfs/main_mkfs.c b/gfs2/mkfs/main_mkfs.c
> > > index 4751f19..d68cf98 100644
> > > --- a/gfs2/mkfs/main_mkfs.c
> > > +++ b/gfs2/mkfs/main_mkfs.c
> > > @@ -456,8 +456,11 @@ fail:
> > >  
> > >  static void are_you_sure(struct gfs2_sbd *sdp)
> > >  {
> > > -	char input[32];
> > > +	char *line = NULL;
> > > +	size_t len = 0;
> > >  	int fd;
> > > +	int ret = -1;
> > > +	int res = 0;
> > >  
> > >  	fd = open(sdp->device_name, O_RDONLY|O_CLOEXEC);
> > >  	if (fd < 0)
> > > @@ -465,14 +468,25 @@ static void are_you_sure(struct gfs2_sbd *sdp)
> > >  	printf( _("This will destroy any data on %s.\n"), sdp->device_name);
> > >  	check_dev_content(sdp->device_name);
> > >  	close(fd);
> > > -	printf( _("\nAre you sure you want to proceed? [y/n] "));
> > > -	if(!fgets(input, 32, stdin))
> > > -		die( _("unable to read from stdin\n"));
> > > +	
> > > +	do{
> > > +		printf( _("\nAre you sure you want to proceed? [y/n]"));
> > This means that the translator needs to know that rpmatch is being used
> > and to also know how to translate this accordingly. Why not use
> > nl_langinfo() to put the response strings into this message?
> > 
> I was looking how nl_langinfo() works, and all it does to help with yes/no 
> questions is to return a regex expression (like [^[SsyY].*) to be used to match 
> the user's answer, it does not return a string or char value like yes/no (en_US) or 
> sim/nao(pt_BR) we can use to replace the y/n in the question string.
> 
> The rpmatch() function takes advantage of nl_langinfo() internaly to match the
> customer's answer according to which l10n the system is set.
> 
> This way, even using nl_langinfo(), the translator person will still need to translate
> the y/n to s/n, since is not possible to use nl_langinfo to replace these values.
> 
> What we could do, is to create an auxiliar function to check the return of nl_langinfo(),
> and, according with this value, write an specific string (s/n, y/n, etc). But, imho, this
> could lead to an error prone condition during string creation.
> 
> What u think?
> 
> Cheers,

It is is not possible to use glibc for this, then lets not worry about
it too much. I just wanted to be certain that there was not some
standard way to deal with these kinds of translations before we create
our own solution to the problem,

Steve.