Hi, On Sun, Feb 27, 2005 at 04:54:30PM +0100, Denis Barbier wrote: > On Sun, Feb 27, 2005 at 03:28:52PM +0000, Nicolas FRAN??OIS wrote: > > * change back non breaking spaces to '\\ ' in the PO files for man pages. > > I quickly read previous messages but did not realize that this change > was needed. Adding "\\ " is much less natural than a non breaking space > (at least for those who are used to write this glyph), so can you please > explain why this conversion was troublesome? Using non breaking spaces by default is troublesome because all encodings don't support this glyph, and thus the generated PO may be invalid in some cases.

On Sun, Feb 27, 2005 at 06:08:06PM +0100, Denis Barbier wrote: > On Sun, Feb 27, 2005 at 05:12:24PM +0100, Nicolas François wrote: > > Hi, > > > > On Sun, Feb 27, 2005 at 04:54:30PM +0100, Denis Barbier wrote: > > > On Sun, Feb 27, 2005 at 03:28:52PM +0000, Nicolas FRAN??OIS wrote: > > > > * change back non breaking spaces to '\\ ' in the PO files for man pages. > > > > > > I quickly read previous messages but did not realize that this change > > > was needed. Adding "\\ " is much less natural than a non breaking space > > > (at least for those who are used to write this glyph), so can you please > > > explain why this conversion was troublesome? > > > > Using non breaking spaces by default is troublesome because all encodings > > don't support this glyph, and thus the generated PO may be invalid in some > > cases. > > Ok, this is indeed troublesome in msgid, but should seldom happen. > Is there a way to keep "\\ " in msgid but convert to non breaking > space in msgstr? Yes and no. With the default behavior, msgid will contain "\\ ". Then, a french translator may use non breaking space in msgstr if he wants. But, during a gettextization (with an original and translation), msgid and msgstr will be handled the same way (both having "\\ " or non breaking space). I can't see a way to have msgid with "\\ " and msgstr with non breaking spaces in this case

Then, if you really need this, you can cheat a little bit with gettext tools. Or, if original is in English: English doesn't use non breaking spaces (IMHO), English man pages (mostly) use them to indicate that two words are in the same argument, and thus only appear in short macro lines. Thus a simple sed script could do the trick (with the "-o with-non_breaking_spaces=1" option).

Is it OK if we document this, and indicate that a project should always use the same option for all languages/po4a commands. If we remove the option, then we can: * just stay with "\\ " in PO. Translators can always use 0xA0 in their msgstr * do the transliteration if the po_out charset is UTF-8 or latin-1. I never tried it.

On Sun, Feb 27, 2005 at 09:20:44PM +0100, Denis Barbier wrote: > At the moment I have no strong opinion on what should be done, since I > believe that English manual pages should be fixed to not use "\ ". After a bit more thinking, I believe that the main problem is that the master file is detected as being an ASCII file, but pre_trans() transforms input text into non-ASCII, so there is some mismatch here. Everything should work out of the box with -M ISO-8859-1, shouldn't it?

Hello Denis, On Sun, Feb 27, 2005 at 07:23:37PM +0100, Denis Barbier wrote: > I had another look at the "Non breaking spaces in man pages" thread; > being on vacation during this period I could not post. I don't know if you noticed this. Martin would like to release tomorrow. Do you think we should fix this issue before the release.