search

bardsoftware / ganttproject

Official GanttProject repository.
http://ganttproject.biz
GNU General Public License v3.0
856 stars 301 forks source link

GanttProject crash leads to non-recoverable gan file #1504

Open bfendler opened 6 years ago

bfendler commented 6 years ago

Unfortunately, I closed the Java dialog box, but a GanttProject crash lead to a gan file that can continues to cycle upon trying to load. There's no way to load the gan file now. The file seems fine--completely visible in vi and doesn't appear corrupted. I would post it, but can't because of extensive ip.

Also, HouseBuildingSample.gan still loads and can be edited and saved, so the build does not seem corrupted.

This is a very damaging bug because the gan file I've spent two days on is now useless. Fortunately, I can rebuild with a pdf export. This is also challenging to report because there doesn't seem to be sufficient logs and I can't post the gan file.

Perhaps other formats should autosave when the gan file is saved?

Sorry I cannot report more, but I thought I should at minimum make this bug known.

dbarashev commented 6 years ago

  1. GanttProject makes autosave and you can use autosave files for recovering. Read this comment for details: https://help.ganttproject.biz/t/2-8-5-2179-build-does-not-respond-to-any-commands/619/9?u=dmitrybarashev

  2. If Help > Recover does not work, please send us the "broken" file if you find it reasonable. You can send it by email (find it on tab "Contacts" on this page: http://www.ganttproject.biz/about) -- it will be kept confidential.

  3. There are numerous reasons why project structure may become broken, and it is unlikely that crash itself is the reason. If you are running Linux, you can find the log file in your home directory and send it to us -- this may help us to find the reasons of crash.

bfendler commented 6 years ago

Thanks for such a quick reply. I can't send the file due to ip, but I was able to load the gan file via "import." I debugged down to a single task by killing tasks, saving as a gan file, and then re-opening. Every task opened with exception to one that reproduced the bug. If I get the chance, I will send a minimal gan file that reproduces this cycling error.

The only strange thing I noticed about the buggy task was that its end-date was one day ahead of all of the sub-task's end-dates. Perhaps I moved a task around and the super-task (buggy task) did not update? But I was able to save all tasks with exception to that strange dated "super-task."