#2329 SEGV on CANVAS_text

[liuchenyifan]

gnuplot 5.5 (last modified in 9.18) on centos linux 7.7.1908

information below from valgrind:

==18744== Invalid read of size 8
==18744== at 0x7C9F93: CANVAS_text (canvas.trm:660)
==18744== by 0x7F9031: term_end_plot (term.c:551)
==18744== by 0x55E728: do_plot (graphics.c:1130)
==18744== by 0x61F74B: eval_plots (plot2d.c:3612)
==18744== by 0x44BB88: plot_command (command.c:1897)
==18744== by 0x4411AB: command (command.c:659)
==18744== by 0x4411AB: do_line (command.c:429)
==18744== by 0x5CD14B: load_file (misc.c:335)
==18744== by 0x406416: main (plot.c:636)
==18744== Address 0x10 is not stack'd, malloc'd or (recently) free'd
==18744==
==18744==
==18744== Process terminating with default action of signal 11 (SIGSEGV)
==18744== Access not within mapped region at address 0x10
==18744== at 0x7C9F93: CANVAS_text (canvas.trm:660)
==18744== by 0x7F9031: term_end_plot (term.c:551)
==18744== by 0x55E728: do_plot (graphics.c:1130)
==18744== by 0x61F74B: eval_plots (plot2d.c:3612)
==18744== by 0x44BB88: plot_command (command.c:1897)
==18744== by 0x4411AB: command (command.c:659)
==18744== by 0x4411AB: do_line (command.c:429)
==18744== by 0x5CD14B: load_file (misc.c:335)
==18744== by 0x406416: main (plot.c:636)
==18744== If you believe this happened as a result of a stack
==18744== overflow in your program's main thread (unlikely but
==18744== possible), you can try to increase the size of the
==18744== main thread stack using the --main-stacksize= flag.
==18744== The main thread stack size used in this run was 8388608.
==18744==
==18744== HEAP SUMMARY:
==18744== in use at exit: 44,597 bytes in 313 blocks
==18744== total heap usage: 712 allocs, 399 frees, 356,237 bytes allocated
==18744==
==18744== LEAK SUMMARY:
==18744== definitely lost: 0 bytes in 0 blocks
==18744== indirectly lost: 0 bytes in 0 blocks
==18744== possibly lost: 0 bytes in 0 blocks
==18744== still reachable: 44,597 bytes in 313 blocks
==18744== suppressed: 0 bytes in 0 blocks
==18744== Rerun with --leak-check=full to see details of leaked memory
==18744==
==18744== For lists of detected and suppressed errors, rerun with: -s
==18744== ERROR SUMMARY: 1 errors from 1 contexts (suppressed: 0 from 0)

[Ethan Merritt]

I cannot reproduce this. In this possibly the issue is that the script triggers a bunch of "history" queries that can retrieve commands from earlier gnuplot runs. So the command causing the dump you show is not actually in the reproducer script, it is retrieved from an earlier session. Maybe. Anyhow for me it runs with no error.

[liuchenyifan]

I'm pretty sure there is only one gnuplot version in there. I'm new for testing gnuplot.
Maybe the enviroment differs.
The version is gnuplot 5.5 (last modified in 9.18) on centos linux 7.7.1908;
The command is "gnuplot bug6"

[Ethan Merritt]

The error in your valgrind output is in the canvas terminal. But when I run it here the script never calls the canvas terminal at all. What it does have is many lines beginning with"hi", which is short for "history". This looks back through commands saved in ~/.gnuplot.history during earlier runs. Maybe an earlier run issued a command that selected the canvas terminal I don't know - that's just a guess. If that's the case, you should probably configure the program to not use the history file configure --disable-history-file

Anyhow the script runs here without generating a fault.

[liuchenyifan]

I try it for more, and still can be reproduced. I have no idea.
But it's a good news if these crashes can not affect gnuplot. Let's leave these alone.
I'll keep testing gnuplot.Thanks anyway!