Version: (using KDE KDE 3.5.1) Installed from: Compiled From Sources OS: Linux Attempting to fit the simpledata.dat file of the Kst tutorial with a second order polynomial crashes Kst. Paste the following into a data file -20 100 -19 90.25 -18 81 -17 72.25 -16 64 -15 56.25 -14 49 -13 42.25 -12 36 -11 30.25 -10 25 -9 20.25 -8 16 -7 12.25 -6 9 -5 6.25 -4 4 -3 2.25 -2 1 -1 0.25 0 0 1 0.25 2 1 3 2.25 4 4 5 6.25 6 9 7 12.25 8 16 9 20.25 10 25 11 30.25 12 36 13 42.25 14 49 15 56.25 16 64 17 72.25 18 81 19 90.25 20 100 ... and plot it with the first column as x and the second as y. Then attempt a 2nd order polynomial fit. Kst crashes for me (and I've just make a fresh build from trunk). No backtrace, but here's what Kst spits out before it SIGABRTs creating KstVector Anonymous Vector 5 creating KstVector Anonymous Vector 6 kst: Update timer 200 kst: Forced update! kst: FATAL: Thread -1212535904 tried to unlock KstRWLock 0x8f05c24 (unlocked) without holding the lock KCrash: Application 'kst' crashing... QSocketNotifier: invalid socket 9 and type 'Read', disabling... QSocketNotifier: invalid socket 4 and type 'Read', disabling... QSocketNotifier: invalid socket 7 and type 'Read', disabling...
Trunk seems to work for me (?) cbn On Thursday 12 October 2006 18:41, Duncan Hanson wrote: [bugs.kde.org quoted mail]
I can't reproduce with trunk either.
I don't seem to have the problem in 1.3 (branch) either...
Can't reproduce.
Do you do the fit using the Data->New Menu, or using the Fit context menu item in the plot? The crash only occurs for me using the Data->New Plugin route. Duncan. On Monday 16 October 2006 09:09, Adam Treat wrote: [bugs.kde.org quoted mail]
I see it. Thanks Duncan.
This deadlock appears for all fit plugins created through the top-level menu OR the datamanager->new plugin manner. It seems to be caused by commit #576228. The commit log says 576228 was a fix for another deadlock. Removing this change fixes this bug, but I don't know what the other deadlock was about, so I'm loathe to revert.
This won't be fixed in 1.3.1 and I'm hesitant to fix it in 1.3.x at all. This deadlock is the result of a fix for another deadlock. The reason for this mess is somewhat complicated. First, automatically adding scalars to the global list is an issue when we need to lock an object that we are providing while we are locked. It can deadlock with the update thread. The "fix" for this caused the regression referenced in this bug. The function it happened in doesn't even belong in KstPlugin. KstScalar's constructor is a mess already, especially with the locking argument, and adding another is not a good answer. I think we need to fix scalars properly. This is a Kst 1.4 item only.
Why does it work from the 2dplot rmb, but not from the dialog? On Tuesday 17 October 2006 23:38, George Staikos wrote: [bugs.kde.org quoted mail]
Different locking/nesting situation.
Note: This is fixed in the hierarchy branch with commit #612802
Duncan, could you test this and confirm that the problem has been fixed. If so please mark this bug report as fixed.
*** Bug has been marked as fixed ***.