Segfault on start when built against ncurses-6.0

[NiLuJe]

Observed at least on x86_64 & armv7, the linked output is from the x86_64 box.

(That's a build from the current master, and ncurses ABI 6, FWIW).

htop 1.0.3 aborting. Please report bug at http://hisham.hm/htop

 Please include in your report the following backtrace: 
htop(CRT_handleSIGSEGV+0x35)[0x41765a]
/lib64/libc.so.6(+0x33440)[0x72391429c440]
/lib64/libncursesw.so.6(+0x31597)[0x723914930597]
/lib64/libncursesw.so.6(+0x32bdd)[0x723914931bdd]
/lib64/libncursesw.so.6(doupdate_sp+0x155)[0x7239149326ce]
/lib64/libncursesw.so.6(wrefresh+0x55)[0x723914923faa]
/lib64/libncursesw.so.6(+0x1e680)[0x72391491d680]
/lib64/libncursesw.so.6(_nc_wgetch+0x2ce)[0x72391491dd91]
/lib64/libncursesw.so.6(wgetch+0x2c)[0x72391491e291]
htop(ScreenManager_run+0xc1)[0x40efa6]
htop(main+0x1d8)[0x40a49f]
/lib64/libc.so.6(__libc_start_main+0xf0)[0x7239142897b0]
htop(_start+0x29)[0x406739]

Additionally, in order to make the above backtrace useful,
please also run the following command to generate a disassembly of your binary:

   objdump -d `which htop` > ~/htop.objdump

and then attach the file ~/htop.objdump to your bug report.

Thank you for helping to improve htop!

[1]    9276 abort      htop

What gdb has to say:

Program received signal SIGSEGV, Segmentation fault.
TransformLine (sp=sp@entry=0x118b660, lineno=lineno@entry=8) at /var/tmp/paludis/sys-libs-ncurses-6.0/work/ncurses-6.0/ncurses/tty/tty_update.c:1261
1261    /var/tmp/paludis/sys-libs-ncurses-6.0/work/ncurses-6.0/ncurses/tty/tty_update.c: No such file or directory.
#0  TransformLine (sp=sp@entry=0x118b660, lineno=lineno@entry=8) at /var/tmp/paludis/sys-libs-ncurses-6.0/work/ncurses-6.0/ncurses/tty/tty_update.c:1261
        oldPair = 0
        newPair = -1708163072
        firstChar = <optimized out>
        oLastChar = <optimized out>
        nLastChar = <optimized out>
        newLine = 0x11a0460
        oldLine = 0x11e8de0
        n = 22
        attrchanged = false
#1  0x00006f9f99eb9bdd in ClrUpdate (sp=sp@entry=0x118b660) at /var/tmp/paludis/sys-libs-ncurses-6.0/work/ncurses-6.0/ncurses/tty/tty_update.c:1073
        i = 8
        blank = {attr = 0, chars = L" \000\000\000", ext_color = 0}
        nonempty = 39
#2  0x00006f9f99eba6ce in doupdate_sp (sp=sp@entry=0x118b660) at /var/tmp/paludis/sys-libs-ncurses-6.0/work/ncurses-6.0/ncurses/tty/tty_update.c:925
        i = <optimized out>
        nonempty = 0
#3  0x00006f9f99eabfaa in wrefresh (win=win@entry=0x1222640) at /var/tmp/paludis/sys-libs-ncurses-6.0/work/ncurses-6.0/ncurses/base/lib_refresh.c:65
        code = <optimized out>
        sp = <optimized out>
#4  0x00006f9f99ea5680 in recur_wrefresh (win=win@entry=0x1222640) at /var/tmp/paludis/sys-libs-ncurses-6.0/work/ncurses-6.0/ncurses/base/lib_getch.c:384
No locals.
#5  0x00006f9f99ea5d91 in _nc_wgetch (win=win@entry=0x1222640, result=result@entry=0x7cf235fb3734, use_meta=0) at /var/tmp/paludis/sys-libs-ncurses-6.0/work/ncurses-6.0/ncurses/base/lib_getch.c:491
        sp = 0x118b660
        ch = <optimized out>
        rc = 0
#6  0x00006f9f99ea6291 in wgetch (win=0x1222640) at /var/tmp/paludis/sys-libs-ncurses-6.0/work/ncurses-6.0/ncurses/base/lib_getch.c:670
        code = <optimized out>
        value = 0
#7  0x000000000040efa6 in ScreenManager_run (this=this@entry=0x1272a40, lastFocus=lastFocus@entry=0x0, lastKey=lastKey@entry=0x0) at ScreenManager.c:191
        prevCh = -1
        result = <optimized out>
        quit = false
        focus = 0
        panelFocus = 0x126ea30
        oldTime = 14395879732.719379
        ch = -1
        closeTimeout = 0
        timedOut = true
        redraw = true
        rescan = false
        sortTimeout = 1
#8  0x000000000040a49f in main (argc=<optimized out>, argv=<optimized out>) at htop.c:234
        lc_ctype = <optimized out>
        flags = {pidWhiteList = 0x0, userId = 4294967295, sortKey = 0, delay = -1, useColors = true}
        ut = 0x11894f0
        pl = 0x11895e0
        settings = 0x118a020
        header = 0x118a6a0
        panel = <optimized out>
        state = {settings = 0x118a020, ut = 0x11894f0, pl = 0x11895e0, panel = 0x126ea30, header = 0x118a6a0}
        scr = 0x1272a40

And here's the objdump of the binary.

[NiLuJe]

Building against ncurses 6.0 in ABI 5 mode appears to work, on the other hand ;).

[hishamhm]

Ouch, I have no idea what could be causing this... header mismatch, perhaps? ncurses.h is known for having lots of macros... I'm running ncurses 5.9 here.

[hishamhm]

I just tested ncurses 6.0 here (in ABI 6 mode, including implementing mouse wheel support, yay!) and had no crashes. Please double check that you're not building using ncurses 5 headers...

[NiLuJe]

I'll double check tomorrow if there wasn't a pkg-config mishap, but that seems unlikely, the native x86_64 build was on Gentoo through portage, and the armv7 build is cross-compiled in a pristine sysroot ;?.

GCC 4.9.3 in both cases, FWIW.

[hishamhm]

Thank you! Please test against the current master; I added support for -lncursesw6 in configure.ac.

[hishamhm]

I managed to reproduce the bug here! Turns out I was building with ABI 6 but accidentally running htop in non-Unicode mode. On to fix it!

[NiLuJe]

Working fine on my end too, thanks! 👍

[NiLuJe]

Err, mostly, that is ^^.

I'm seeing a weird graphical issue where sometimes, the coloring of a commandline will go a bit haywire. The weird colors shift around when resizing the window, and eventually settle down to something normal. So far, I've only seen this on very long shadowed (different user) commandlines.

x86

armv7

... after a resize

... and okay after a second resize.

I can create a separate issue for this if you prefer?

EDIT: The "Shadow other users' processes" option is irrelevant, the disruption is still the same, just with a different set of colors ;).

[hishamhm]

Oh, thanks for the screenshots!
Please open a new issue, so that other users find it if they bump into it.