htop 3.1.1 segfaults when adding "CPU average" meter

[scalisi]

1. I installed htop 3.1.1 from source on Debian Buster.
2. I ran htop and noticed that my custom header had reverted to the default.
3. I deleted my htoprc.
4. I reconfigured htop, which included adding the CPU average meter to the header.
5. I quit htop.
6. "Segmentation fault" was printed in my terminal.
7. I ran htop again and saw that my custom header had reverted to the default again.
8. After some trial and error, it seemed that the segfault occurred only when the CPU average meter was added to the header.
9. I deleted my htoprc again and the only change I made was adding the CPU average meter to the header.
10. I quit htop.
11. "Segmentation fault" was printed in my terminal.

I didn't experience this issue with htop 3.1.0, which I also installed from source on Debian Buster.

[fasterit]

Could you please compile with --enable-debug¸ and then provide a gdb stack backtrace of the segfault?

gdb -batch -ex "run" -ex "bt full" ./htop 2>&1 should do.

NB: DLange could not reproduce locally.

[scalisi]

Program received signal SIGSEGV, Segmentation fault.
__GI___libc_free (mem=0x31) at malloc.c:3093
3093    malloc.c: No such file or directory.
#0  __GI___libc_free (mem=0x31) at malloc.c:3093
        ar_ptr = <optimized out>
        p = <optimized out>
        hook = 0x0
#1  0x000055555557d555 in String_freeArray (s=0x5555555a5c80) at XUtils.c:158
        i = <optimized out>
#2  0x000055555557a49f in Settings_delete (this=this@entry=0x5555555a58d0) at Settings.c:31
        i = 0
#3  0x0000555555567bcc in CommandLine_run (name=<optimized out>, argc=<optimized out>, argv=<optimized out>) at CommandLine.c:395
        lc_ctype = <optimized out>
        flags = {pidMatchList = 0x0, commFilter = <optimized out>, userId = <optimized out>, sortKey = <optimized out>, delay = -1, useColors = true, enableMouse = <optimized out>, treeView = <optimized out>, allowUnicode = true, highlightChanges = <optimized out>, highlightDelaySecs = <optimized out>, readonly = <optimized out>}
        ut = 0x5555555a0930
        dc = 0x5555555a0ac0
        dm = 0x0
        pl = 0x5555555a0c30
        settings = 0x5555555a58d0
        header = 0x5555555a5d70
        panel = <optimized out>
        state = {settings = 0x5555555a58d0, ut = 0x5555555a0930, pl = 0x5555555a0c30, mainPanel = 0x5555556c2d10, header = 0x5555555a5d70, pauseProcessUpdate = false, hideProcessSelection = false}
        scr = 0x5555556c6b60
#4  0x00007ffff7c2909b in __libc_start_main (main=0x555555563a70 <main>, argc=1, argv=0x7fffffffdeb8, init=<optimized out>, fini=<optimized out>, rtld_fini=<optimized out>, stack_end=0x7fffffffdea8) at ../csu/libc-start.c:308
        self = <optimized out>
        result = <optimized out>
        unwind_buf = {cancel_jmp_buf = {{jmp_buf = {0, -3130608271126653967, 93824992295568, 140737488346800, 0, 0, -9090360737828137999, -9090342774506613775}, mask_was_saved = 0}}, priv = {pad = {0x0, 0x0, 0x7fffffffdec8, 0x7ffff7ffe190}, data = {prev = 0x0, cleanup = 0x0, canceltype = -8504}}}
        not_first_call = <optimized out>
#5  0x0000555555563aba in _start ()
No symbol table info available.

[scalisi]

I spent more time configuring htop and noticed that it segfaults when adding certain other meters, not just CPU average.

[BenBE]

Can you provide an example configuration that causes the crash?
In addition could you compile htop with ASAN (-fsanitize=address) enabled? This looks like some uninitialized memory that gets accessed when reading the configuration fails.

[cgzones]

p.s.: the allocation became an issue with 2a13e15 free'ing the array via String_freeArray().

[BenBE]

I figured from your patch, but actually it was problematic before that change in a different way: The original code already checked for the length of that array to determine how many modes it had to read. That was what caused the change of my patch, as the length field was used somewhat inconsistently.