Hi. In gnuplot4 the following script worked as expected:
set output
set view map
splot '-' matrix using 1:2:3 notitle with image
0.0 1.0 2.0 3.0 4.0
5.0 6.0 7.0 8.0 9.0
5.0 6.0 7.0 8.0 9.0
e
This would show an image plot. In the gnuplot built from the most recent sources, however, this doesn't work. I get an error:
warning: Too many axis ticks requested (>inf)
The plot window pops up, but then immediately goes away.
No problem appeared when I tried to reproduce this here. Could this be the consequence of some background setting, e.g. in your ~/.gnuplot or similar?
Hi. Thanks for checking. I just ran a few more experiments and have more information. I use a tiling window manager, so new X windows are resized immediately after they pop up. If I allow the original window size to be retained, then things work ok. It is this window resize that makes gnuplot confused. I see issues with both the x11 and the qt terminals, and as I said, this used to work previously.
Thanks
I get a segfault, and the code where it segfaults is clearly incorrect. It was written to handle 2D plot structures, but in this case it ends up being executed to update a 3D plot structure. The pointers don't match. I have "fixed" this in the attached patch in the sense that both 2D and 3D plot structures are now correctly indexed, but I worry that the real problem is that this code path should never be entered for 3D plots at all.
I haven't applied this patch to CVS yet because I would like to understand better where and how the regression crept in. This test script acts a little oddly with version 5.0.2 but does not segfault. That is strange, because the same invalid code is present in 5.0
Anyhow, please give the patch a try. I think it will in fact fix this particular problem case, but I am less sure it is correct for all cases (e.g. true 3D images rather than 'set view map').
Last edit: Ethan Merritt 2016-01-17
Fixed for 5.1
Unfortunately I don't see how to fix it for 5.0 short of back-porting the revised image dimension tracking code added to 5.1 (May 2015 and Jun 2015). That's a larger change than I usually consider for a patchlevel update, but since it deals with the present bug as well as the earlier ones (e.g. #1607) maybe we should apply it for 5.0.3
The new code has been backported and will be in 5.0.3