Tails

#1 Updated by anonym 2014-11-18 17:12:40

The following quick’n’dirty fix results in a bootable Jessie-based Tails USB installation:

--- /usr/lib/python2.7/dist-packages/liveusb/creator.py.orig    2014-11-18 16:27:31.056000000 +0000
+++ /usr/lib/python2.7/dist-packages/liveusb/creator.py 2014-11-18 16:36:58.204000000 +0000
@@ -250,12 +250,12 @@
                + [ self.drive['parent'] ])
         self.popen(cmd, shell=False)

-    def _gdisk_script(self, gdisk_cmds):
+    def _gdisk_script(self, gdisk_cmds, passive=False):
         cmd = [ '/sbin/gdisk', self.drive['device'] ]
         stdin = tempfile.TemporaryFile(mode='w')
         stdin.write('\n'.join(gdisk_cmds) + '\n')
         stdin.seek(0)
-        self.popen(cmd, shell=False, stdin=stdin)
+        self.popen(cmd, passive=passive, shell=False, stdin=stdin)

     def create_hybrid_mbr(self):
         # only relevant with OS that don't know how to access partitions on GPT
@@ -268,14 +268,14 @@
                              'N',   # no protective partition
                              'w',   # write table to disk and exit
                              'Y',   # proceed, possibly destroying data
-                           ])
+                           ], passive = True)

     def remove_hybrid_mbr(self):
         self._gdisk_script([ 'x',   # extra functionality
                              'n',   # create a new protective MBR
                              'w',   # write table to disk and exit
                              'Y',   # proceed, possibly destroying data
-                           ])
+                           ], passive = True)

     def is_device_big_enough(self):
         if self.system_partition_size * 1024**2 <= self.drive['size']:

#3 Updated by intrigeri 2014-11-25 20:44:00

Looks like a bug in gdisk, possibly the same as http://sourceforge.net/p/gptfdisk/mailman/message/32068566/. Would be good to have this fixed in Jessie. Next step is to check with upstream if it’s a known issue (and possibly to try reverting the commit pointed at by the message linked above).

#4 Updated by anonym 2014-12-02 16:56:44

I think you are right. Upstream commit a920398 says: ‘Fixed bug that caused spurious “1” exit condition in gdisk.’ Curiously, that commit is from March 29, and there has been no release since (in facts, it’s master’s HEAD). Hmm.

Also, the commit is non-atomic, but the following extract of it hopefully fixes this issue (haven’t tested):

--- a/gdisk.cc
+++ b/gdisk.cc
@@ -15,6 +15,7 @@ int main(int argc, char* argv[]) {
    GPTDataTextUI theGPT;
    string device = "";
    UnicodeString uString;
+   int isError = 0;

 #ifndef EFI
    cout << "GPT fdisk (gdisk) version " << GPTFDISK_VERSION << "\n\n";
@@ -49,16 +50,20 @@ int main(int argc, char* argv[]) {
             device = (string) argv[1];
          } else { // 3 arguments, but none is "-l"
             cerr << "Usage: " << argv[0] << " [-l] device_file\n";
+            isError = 1;
          } // if/elseif/else
          if (device != "") {
             theGPT.JustLooking();
             if (theGPT.LoadPartitions(device))
                theGPT.DisplayGPTData();
+            else
+               isError = 1;
          } // if
          break;
       default:
          cerr << "Usage: " << argv[0] << " [-l] device_file\n";
+         isError = 1;
          break;
    } // switch
-   return 1 ;
+   return (isError);
 } // main