5:27 p.m.
Hello,
There is a first issue which causes a bad formatting of the po4a.7 and
Locale::Po4a::TransTractor manpages:
Here is an example:
http://po4a.alioth.debian.org/es/Locale/Po4a/TransTractor.html#DESCRIPCION
(at the end of this section, the "bone" is broken)
This is caused by a line ending by a \ in a no-wrap paragraph.
The Po unescape function replace ([^\\])\\n by $1\n
I think it should instead replace "even number of \ followed by \n"
thus s/([^\\](\\\\)*)\\n/$1\n/g;
(I'm joining Po.pm.escape.patch)
If you think it is the right way to fix this issue, then I will also
change other similar regular expressions in this module (there are many
others).
The second issue is one of the problems detected by the WDIFF category of
the testsuite (some text is present in the original, but not in the
translation, or the contrary).
If a man page contains something like that (in a wrapped paragraph):
\fP
foo
then the font stack may "delay" the font modifier, and po4a generates the
following paragraph: "\nfoo". The result of gettext with this msgid is
always a null string.
I will fix the Man module so that the empty line will be pushed without
translation, but the Po module may need the second patch.
(I'm joining Po.pm.defined_empty.patch and foo.1, which trigger this
issue)
Both patches work, but I'm not sure of my perl and I would prefer you have
a look at them.
Regards,
--
Nekral
10:45 p.m.
On Thu, Nov 25, 2004 at 12:27:46AM +0100, Nicolas François wrote:
The Po unescape function replace ([^\\])\\n by $1\n
I think it should instead replace "even number of \ followed by \n"
thus s/([^\\](\\\\)*)\\n/$1\n/g;
(I'm joining Po.pm.escape.patch)
If you think it is the right way to fix this issue, then I will also
change other similar regular expressions in this module (there are many
others).
Yes, I think so. At least, I cannot think of any better solution. Simply,
make sure to add some comments around the regexp since it becomes quite
hairly.
The second issue is one of the problems detected by the WDIFF
category of
the testsuite (some text is present in the original, but not in the
translation, or the contrary).
If a man page contains something like that (in a wrapped paragraph):
\fP
foo
then the font stack may "delay" the font modifier, and po4a generates the
following paragraph: "\nfoo". The result of gettext with this msgid is
always a null string.
This is a good ol' bug of testing the string instead of testing the
existance of the string. The little brother of testing the string instead of
testing its length. Your fix is exactly what we need, thanks a lot.
I will fix the Man module so that the empty line will be pushed
without
translation, but the Po module may need the second patch.
Yeah, kill that beast, please.
Both patches work, but I'm not sure of my perl and I would prefer
you have
a look at them.
Well, not much to say about one liner patches. They look good. As I said,
they just miss some comments.
Thank you very much,
Mt.
6:54 p.m.
Hello,
One of the patches was uploaded. The other one (\n preceded by an even
number of \) is more complicated.
This issue appears in unescape_text (\\n is not tranlated to an end of
line, which causes formatting issues) and in quote_text (in no-wrap mode,
no newline is added after a \\n, but I think it has no implication on the
output document).
There may also be the same issue in unquote_text, but I could not find a
way to trigger it.
My solution is maybe too complicated:
+ # unescape newlines
+ # NOTE on \G:
+ # The following regular expression introduce newlines.
+ # Thus, ^ doesn't match all beginnings of lines.
+ # \G is a zero-width assertion that matches the position
+ # of the previous substitution with s///g. As every
+ # substitution ends by a newline, it always matches a
+ # position just after a newline.
+ $text =~ s/( # $1:
+ (\G|[^\\]) # beginning of the line or any char
+ # different from '\'
+ (\\\\)* # followed by any even number of '\'
+ )\\n # and followed by an escaped newline
+ /$1\n/sgx; # single string, match globally, allow comments
+ # unescape tabulations
+ $text =~ s/( # $1:
+ (^|[^\\]) # beginning of the line or any char
+ # different from '\'
+ (\\\\)* # followed by any even number of '\'
+ )\\t # and followed by an escaped tabulation
+ /$1\t/mgx; # multilines string, match globally, allow comments
Do you have another idea to avoid using (and documenting) \G?
In canonize, I don't thing it is needed to protect \n against a preceding \,
and I'm wondering if
$text =~ s/([^\\])\n/$1 /gm;
$text =~ s/ \n/ /gm;
$text =~ s/([^\\])\n/$1 /gm;
should not simply be changed to:
$text =~ s/\n/ /gm if ($text ne "\n");
(do not do anything if ($text eq "\n"), because it mess up the first
string - the header)
Regards,
--
Nekral
7098
days inactive
7115
days old
2 comments
2 participants
participants (2)
-
Martin Quinson -
Nicolas François