search

kaitz / paq8pxd

GNU General Public License v2.0
68 stars 7 forks source link

Bug report: SZDD misreconstruction #10

Closed M-Gonzalo closed 3 years ago

M-Gonzalo commented 4 years ago

I found a file (possibly damaged) that doesn't restore losslessly

ati3duag.dl_.zip

EDIT: This is NOT the correct file; see the update.

kaitz commented 4 years ago

I compared decompression. Reported identical. I extracted and compared two files, file hashes were same. What option was used, cpu arch?

M-Gonzalo commented 4 years ago

You know, I had to go through the entire testbed again because I somehow swapped the original file with the reconstructed version so of course nobody could reproduce the problem. Sorry about that... -facepalm-

The origin machine was a Google colab accessed from a Jupyter notebook so I can't know for sure the exact specs, but it probably was an Ubuntu 16.04 x86-64 with ~16gb RAM and an Intel Xeon. My machine is a Manjaro Linux up-to-date, Intel Celeron, 4 gb RAM. I used g++ 10.2 to compile paq8pxd and I included a copy of my binary just in case on the attachment, as well as a tiny bash script called error.sh, which should output the following:

1203810 ati3duag.dl_original
1206972 ati3duag.dl_extracted
2354127 ati3duag.dl_.paq8pxd89

...and the output of paq8pxd is logged to comp.log and dec.log

On the upside, this is only one of several thousands of files I checked so you can probably say paq8pxd pre-processing is pretty stable, bar the occasional misfire.

ati3duag.dl_.zip

kaitz commented 4 years ago

7z also fails on this file, about 850kb is extracted. If Test button is used then archiver updates file size to 543xxx and its allmoust the same as pxd fails. :) Will add to my todo list.

kaitz commented 3 years ago

See v91