Gitx GitX

Comments and changes to this ticket

• 

  Johannes Gilger September 9th, 2009 @ 12:59 PM

  • State changed from “new” to “open”

  Yes, we're still having some problems with that. I guess the following is the case (please confirm or deny):

  1. You're staging using the stage-lines/stage-hunk buttons (and not staging file-wise) right?
     
  2. The hunk/lines you're staging contain some kind of weird characters, like umlauts, UTF-8 stuff, whatever (screenshot of the hunk would be nice).

  What's strange is that these message keep appearing after you close GitX

• 

  Pedro Melo September 9th, 2009 @ 01:13 PM

  Hi,

  1. Yes, only when I use stage-hunk/stage-line buttons;
  2. This file is Latin1-encoded, and I can only notice the º used in there (http://www.simplicidade.org/share/gitx_commit_prob.png).

  Two extra points:

  1. please ignore the messages keep on appearing after GitX exits, I was mistaken;
  2. I had this error message with git gui before, but in this case, git gui works fine.

  Best regards,

• 

  Pieter de Bie September 9th, 2009 @ 01:23 PM

  Ah, that explains it then.

  I'll describe why this fails, but I think Johannes has already discovered this.

  When reading the patch, the stuff goes through PBEasyPipe. Then, on
  line 93, we try to convert the data to a usable NSString. This happens
  like this:

  NSString *string = [[NSString alloc] initWithData:data
  

  encoding:NSUTF8StringEncoding];
  

  if (!string)
      string = [[NSString alloc] initWithData:data
  

  encoding:NSISOLatin1StringEncoding];
  

  So, we first try to create a UTF8 string from it. If that fails, we
  try it as a Latin1 string. In your case, we cane the latin-1 string,
  because the º won't encode to UTF8 properly.

  This shouldn't be that much of a problem, until we try to apply the
  patch. Again this stuff goes through PBEasyPipe, but the data is
  written a bit higher in the code (line 86):

      [inHandle writeData:[input dataUsingEncoding:NSUTF8StringEncoding]];
  

  So, suddenly we write the º character as UTF8, instead of Latin-1.
  Git, having no clue about encodings at all, just sees a different byte
  sequence, and says the patch cannot be applied.

  So basically there are two problems here.

  1) GitX dumps the encoding information of the file. We should probably
  store that somewhere in the PBGitFile or whatever we use, to make sure
  we can restore it. We then have to extend PBEasyPipe to be able to
  output Latin-1 stuff.

  2) GitX doesn't report failed patches, so you don't have a clue that
  it didn't apply.

  Both shouldn't be that hard to fix, though the first is a bit
  annoying. I'm a bit tired of all the encoding madness :). The second
  especially should be easy to fix.

• 

  Pedro Melo September 9th, 2009 @ 01:40 PM

  Forgive my complete noobness regarding Cocoa and GitX, but couldn't you just treat it all as a byte sequence, and convert to NSString only when you need to display it?

  Putting it another way: everything read from git and sent to git is treated as a binary blob, and only gets "upgraded" to a NSString for display purposes.

  Best regards,