File system in Sidebar does not display directory names that are upper ascii (nor double-byte). Create a directory named "téstíñg" Click on File System on Sidebar and choose the drive where the directory is. Note that the directory name is now "testing"
I am using nsFileSpec, which only has a single-byte interface to get the file names. Reassigning to John: if you make this double-byte, then I can display int'l chars. FWIW, I believe that the file system data source is "out", so I'm going to change the target milestone on this to something WAY in the future.
OK, so it's academic. But I'd like some pointers on exactly what's required. Is calling nsString::GetUnicodeString() (or whatever it's called) enough? If so, you can do it outside of nsFileSpec. I note that all the viewer clients of nsFileSpec do this extra step themselves (see, for example, nsBrowserWindow.cpp).
Filenames used by the OS' filesystem are encoded in charset encodings specific to each localized OS. On Mac: most European filenames are in MacRoman, Japanese filenames in Shift-JIS. On Win9x: Euro filename are in MS-Latin1, Japanese filenames are in Shift-JIS. On Unix: Euro filenames are in ISO-Latin1, most Japanese filenames are in EUC-JP. So we need to determine which encoding is being used by the file system, so the proper to/from Unicode conversion can be performed before reading or writing filenames. nsString are 2-byte Unicode (w/a special case to allow 1-byte (7-bit) ASCII). So if nsFileSpec returns nsString, then it will need to do the proper conversion and not the caller of nsFileSpec. This will also hide the platform dependency of the filesystem encodings.
Adding firstname.lastname@example.org to cc list.
Dougt should own this now.
No, you need to convert char* to PRUnichar* by using Unicode Converter. Find the nsIPlatformCharset service and ask for the charset of the FileName selector. Then ask Character Converter Manager for an nsIUnicodeDecoder for it. Then you convert it from char* to PRUnichar*. Naoki have change several code in widget to make the nsIFileWidget work. You may look at those code to see how he do that. He convert PRUnichar* to char* by using nsIUnicodeEncoder, and you just need the reverse one by using nsIUnicodeDecoder.
The side bar should be using nsIFile. If it does, strings returned by GetPath() are in UTF8.
Where is the code which convert the file system text into UTF8 in nsIFile ? rjc- yesterday you said you are dealing this this bug ?
Why the method GetPath() passing char* instead of PRUnichar* / nsString ? We should always passing PRUnichar* or nsString in the interface for any text which may contain non ASCII. The interface problem is bigger than the implementation problem. Once the interface is wrong, you need to spend triple time to fix the problem.
ftang: No, this isn't the bug I was referring to.
nsIFile return strings in utf8. If I create a file called téstíñg, and use an ls tool that uses nsIFile, I the path returned is téstíñg. This is not a nsIFile bug. I am reassiging to waterson. Need to use nsIFile. (if I am missing something, please let me know)
>nsIFile return strings in utf8. If I create a file called téstíñg, and use >an ls tool that uses nsIFile, I the path returned is téstíñg. >This is not a nsIFile bug. >I am reassiging to waterson. Need to use nsIFile. This sound exactly like a nsIfile problem.... 1. the file system does not use UTF8 as encoding for the file system, so you should NOT return UTF8 unless you do explicit conversion between the file system charset and UTF-8 2. the file: url historially use file system charset (not UTF-8 again) in escaped form. To maintain the compatability with the old file: URL, we should treate the %xx escape in file: url as file system charset instead of escaped UTF-8. Therefore, the nsIFile better not use UTF-8 but file system charset.
As waterson's said "File System in Sidebar" is OUT now. I cannot find the way to show the file system in sidebar now.
qa contact to shrir for sidebar bugs
there are like a million better ways to do this now.