Re: "Calculated and written ECMA-119 tree end differ" under very specific circumstances

[Thomas Schmitt]

Hi,

it appears that not a second bug was unveiled but that rather the AAIP
entries need the same counting of the continuation area padding for their
prediction, which spoiled the prediction with ZF entries.

This observation led to a heuristic fix of the fix, which passed all
tests which failed with previous versions.
But i do not really know yet why i must add the padding to the counting
before predicting the storage needs of the AAIP entries. (It must be
hidden in the way how the write pass handles the padding. I trashed half
a dozen theories meanwhile.)

Whatever, i overwrote the dangerous GNU xorriso-1.5.5 tarball and invite
for testing the new one:

  http://www.gnu.org/software/xorriso/xorriso-1.5.5.tar.gz
  MD5 a5e3dc4e0b92be6837f9ed6cbf7f9df5
  Version timestamp :  2021.02.03.115443

Here are the commits since libisofs-1.5.4 so far (for the curious).

The first attempt to fix the problem:
  
https://dev.lovelyhq.com/libburnia/libisofs/commit/058f18d37a928bc5298474848d53fad152fd8d1b

The current fix for the fix:
  
https://dev.lovelyhq.com/libburnia/libisofs/commit/98aea0c18ac6187cf12513107053fdb2086f85f7

I will continue to investigate.


Have a nice day :)

Thomas