The above code was originally tested in Maverick. I noticed that declaring the language array with 100 elements instead of 50 reduces the risk for a segmentation fault. (50 or 500 elements trigger typically errors, while 100 works more safely!?)
Anyway, when I changed from 50 to 100 in the above code, GDM in Maverick seemed to 'almost' make it. The suggested language for added users was taken from the system $LANGUAGE as expected. However, after having typed the password and clicked the button, GDM hanged, and I had to switch off the power to shut down.
As regards GDM in Natty, getenv() seemed to fail in grabbing $LANGUAGE, so the code within the if block was never run, and login worked as usual. Consequently, the suggested language for added users were taken from $LANG - just as before.
The reason why I began to dig into this is that I have started to write an i18n document for the system docs. When I was about to say that the system LANGUAGE (and LC_MESSAGES) variables serve as defaults for added users, it struck me that I ought to test it.
Unless any of you is able to solve the problem easily, I would say that it can wait til 11.10. After all, the most important issue that is mentioned in this bug may be fixed with the linked MP.
The above code was originally tested in Maverick. I noticed that declaring the language array with 100 elements instead of 50 reduces the risk for a segmentation fault. (50 or 500 elements trigger typically errors, while 100 works more safely!?)
Anyway, when I changed from 50 to 100 in the above code, GDM in Maverick seemed to 'almost' make it. The suggested language for added users was taken from the system $LANGUAGE as expected. However, after having typed the password and clicked the button, GDM hanged, and I had to switch off the power to shut down.
As regards GDM in Natty, getenv() seemed to fail in grabbing $LANGUAGE, so the code within the if block was never run, and login worked as usual. Consequently, the suggested language for added users were taken from $LANG - just as before.
The reason why I began to dig into this is that I have started to write an i18n document for the system docs. When I was about to say that the system LANGUAGE (and LC_MESSAGES) variables serve as defaults for added users, it struck me that I ought to test it.
Unless any of you is able to solve the problem easily, I would say that it can wait til 11.10. After all, the most important issue that is mentioned in this bug may be fixed with the linked MP.