Excel document lost contents, became blank

juannm · February 27, 2019, 8:57pm

I’ve been editing an Excel document which only contains plain data (no formulas, no scripting); it gets saved as Excel format (XLS). At some point, the file started showing empty when opened in Calc, although the file size has not been reduced so all actual data must still be there.

I’m assuming a save file bug in Calc caused this issue. It looks similar to this one: https://bugs.documentfoundation.org/show_bug.cgi?id=97597
but that one got solved already for LibreOffice 5.x

Version: LibreOffice Calc 6.1.5.2 on Ubuntu 16.04 (Xenial) and Windows 7.
I don’t have a MS Excel installation so I cannot test if the file opens up in there. XLS format is used for compatibility with other people’s work.

BROKEN file: https://www.dropbox.com/s/iuz08dnk5pjztr5/BROKEN.xls?dl=1
GOOD file: https://www.dropbox.com/s/ok1vgn74u60zget/GOOD.xls?dl=1

Is this really a bug? Is there anything I can do to recover the contents of the file? How to prevent LibreOffice from losing document contents? (seems pretty serious)

erAck · February 27, 2019, 10:11pm

The file does not contain only plain data, it contains graphics objects as well. Seems the Escher graphics stream is broken, the document loads in Gnumeric with cell content but displays bad graphics objects (which are supposed to be logo, pinout, etc.). I’m attaching the file as could be saved with Gnumeric, BROKEN-recovered.xls

Loading the good document in LibreOffice and saving again does not break, you maybe have added something that either can’t be stored in .xls or isn’t handled properly with that file format. The fixed bug you mentioned is unrelated.

As a general advice, do not use the old binary .xls file format as it does have its limitations. Use ODF .ods instead or OOXML .xlsx

If you somehow can reproduce the failure then opening a bug and attaching a sample document and precisely detailed steps to reproduce would be welcome. In any case, opening a bug and attaching the documents can be worthwhile as maybe the reading filter could be enhanced to at least preserve the cell content in this case.

juannm · February 27, 2019, 11:13pm

I’m sorry, those images are indeed logos and a small diagram, they have been there for so long that I didn’t even remember them… I’d like to remove those but it’s not something I can decide. I’ll try to reproduce the issue, and will open a bug report. Thanks for the analysis of the files!

mikekaganski · February 28, 2019, 6:15am

Just in case - this is the broken file saved with Excel 2016.