Backup is always empty #10853
Comments
|
Hm that should not happen. We look into it. |
|
I checked and the backups are there and are non empty. Trying to kill manually jabref to reproduce the bug with unsaved changes and recovering doesn't seem to reproduce the issue, maybe I will wait for a crash to recheck then ... (and maybe open jabref with logs/debug enabled to capture such an event the next time it happens). In the meantime, maybe I should rename the issue to |
|
Do you have any idea which content you put into the entry editor so that JabRef cracked during writing the contents? Could you tey to check the log files of JabRef? There should be an exception recorded. That could give us some hints... |
|
Oh it's not a permanent log file is it ? On 5.12 the log file is overwritten on each jabref boot (so I'm unable to find any bug info as it's gone once I relaunch jabref after a crash). This time I think I have a way to reproduce the bug (with logs as well). I'm kinda sorry for the long wait (and the original issue that was not detailed enough, it's just that these things happen when I use jabref, so when I'm busy, this time I happened to see a pattern) The input jabref failed on is this one (ASCII clipboard content viewed from clipboard viewer) This is copied from the reference title Chaos in Dynamical Systems. seen here Shadowing theorem To reproduce, just paste this clipboard into a new article entry, in the title section, do not save and wait for jabref to crash (occurs in less than a minute). Then, on reboot, the backup will be empty with this error message 
Jabref will also fail to reload the library if loaded at this point. The stacktrace shown in that window is the following : The stacktrace that mac os gives for the crash is the following mac os errror logI've also linked this stacktrace as a txt file for good measure |
|
This crash puts jabref in a state where the backup is not accessible (clicking on |
|
Having a mac here as well, I cannot reproduce this in the latest dev version. https://builds.jabref.org/main/ (Arm suffix is for. mac m1) |
|
I've re tested on the latest dev version and it still happens. The thing is I'm unable to reproduce the bug in a fully reproducible way (as the crash only occurs sometimes even though the chain of action is the same). For now the steps that (sometimes) made jabref crash are the following
If the app crashes upon pasting this, it will do some at most 2 times before doing the same action doesn't seem to trigger the crash anymore on the same library file. I don't know how actionable this information is, I do think there could be more things done to prevent overwriting your perfectly valid bib file by a blank one : I believe the |
|
@Doublonmousse would you be so kind to provide the broken Also, in the screenshot you show, you partially seem to deal with UTF16 encoded text. JabRef bibfiles only supports parsing UTF8 encoded text nowadays. Maybe sometimes you copy and paste UTF16 encoded text into your bib-file? Does that cause the crash? |
I don't really have a non working I still don't have a fully reproducible example either (and I haven't used jabref enough on more recent versions to say whether this is fixed or not). So it's overall very obscure (and when it did crash it wasn't necessarily immediate either). I feel like having some additional cautionary check when using the recovery dialog would be more than enough of a fix (make it so it's not possible to override the current bib file with an empty backup). Everytime it did crash in this way, there wasn't actually a data loss (except maybe for the entry on which it did crash). |
|
Related: #11454 |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
JabRef version
5.12 (latest release)
Operating system
macOS
Details on version and operating system
Sonoma 14.2, M2
Checked with the latest development build (copy version output from About dialog)
Steps to reproduce the behaviour
.bibfile ....bibfileAppendix
EDIT : There is actually a
review backupbutton.I think there should be a warning inviting to review the backup before proceeding if it's empty or small compared to the current
.bibfileIf I hadn't saved my
.bibfile in a local.gitrepository, it would have been erased multiple times by Jabref.This all happened on an external drive
The text was updated successfully, but these errors were encountered: