Viewing a translation page fails in unix2newlines

    * Module zope.tales.tales, line 698, in evaluate
      /srv/launchpad.net/production/launchpad/lib/canonical/launchpad/templates/pofile-translate.pt
      Line 320, Column 24
      Expression: <PythonExpr (messageSet.getTranslation(index))>
         - Names:
            {'args': (),
             'context': <POFile at 0x2aa080d450>,
             'default': <object object at 0x2a955c01d0>,
             'loop': {},
             'nothing': None,
             'options': {},
             'repeat': {},
             'request': <zope.publisher.browser.BrowserRequest instance URL=https://launchpad.net/products/sympa/+series/main/+pots/sympa/hu/+translate>,
             'template': <zope.app.pagetemplate.viewpagetemplatefile.ViewPageTemplateFile object at 0x2a9ecccad0>,
             'usage': <zope.pagetemplate.pagetemplate.TemplateUsage object at 0x2aa18d9a10>,
             'view': <zope.app.pagetemplate.simpleviewclass.SimpleViewClass from /srv/launchpad.net/production/launchpad/lib/canonical/launchpad/templates/pofile-translate.pt object at 0x2aa17e2a90>,
             'views': <zope.app.pagetemplate.viewpagetemplatefile.ViewMapper object at 0x2aa18d9150>}
    * Module zope.tales.pythonexpr, line 59, in __call__
      __traceback_info__: (messageSet.getTranslation(index))
    * Module <string>, line 0, in ?
    * Module canonical.launchpad.browser.pomsgset, line 176, in getTranslation
    * Module canonical.launchpad.helpers, line 829, in unix2windows_newlines
      raise ValueError('\'%r\' is already converted' % text)

ValueError: 'u'A levelez\xe9si lista kezel\u0151 nem tudta tov\xe1bb\xedtani \xd6nnek a levelet, mert nem tudta el\xe9rni a szem\xe9lyes tanus\xedtv\xe1ny\xe1t (e-mail: %1).\r\n'' is already converted

Same for the dutch language

Oddly enough, I can're reproduce these anymore. Can anyone reproduce?

I can't reproduce it too.

It still occurs, we just got a dupe reported. It's kind of hard to reproduce because I think the string order changes -- or something else transient happens.

Ok, so the plan to fix this is:

All translation imports (using web or .po uploads) will be modified so:

- if the translation has '\r\n' but the msgid does not have it, we will remove the '\r' from the translation.
- if the translation has '\n' and the msgid has '\r\n' the translation will have too '\r\n'.
- if the msgid and the translation has '\r\n' nothing will be changed.

Also, when a new .pot is imported, if any msgid has a '\r' char we will warn the uploader because we think is better if the source strings has only '\n' chars as the translator does not have a way to easily see or understand it.

This fix is already implemented and waiting for a review. If all things goes as it should, it will be applied next Tuesday in our production server.

The fix is done. Will be installed in our servers soon.