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

[Thomas Schmitt]

Hi,

Valtteri Vuorikoski wrote:
> I was just getting started with xorriso for
> archiving so I think I can hold for a while too.

You can trust the results if -check_md5_r finds no deviations. This command
reaches out from the directory records to the data file content and to
the independently stored array of MD5 sums. If they match, then the file
block addresses in the directory record are correct.

But if it goes wrong on BD-R, then the medium might be spoiled for any
further use.
You could let xorriso write the ISO into a data file and verify it
before burning the ISO image onto BD-R by

  xorriso -as cdrecord -v dev=/dev/sr0 "$iso_image"

("-as cdrecord" options are documented in man xorrecord.)


> Sorry I can't be of much use in tracking down the bug.

It's a matter of code study, experiments, debugging, and banging the
head on the table.

I really wonder how the obviously wrong handling of contination area
padding with zisofs in libisofs <= 1.5.4 can hide some not-so-rare
occasion for the opposite mistake.

Meanwhile i reduced the number of needed files to 585 in a single
directory. The 100 in my tests of yesterday seem not to be enough.


Have a nice day :)

Thomas