loop on notcurses-demo fails to write JSON about 1% of the time #407
Comments
|
Oddly enough, this seems to happen most often at two widths: 104 and 174. whaaaa |
|
I've currently got a hot reproducible segfault in |
|
|
|
It's actually in ==2315424== Process terminating with default action of signal 11 (SIGSEGV)
==2315424== at 0x48CDF8F: raise (raise.c:46)
==2315424== by 0x48CE10F: ??? (in /lib/x86_64-linux-gnu/libpthread-2.30.so)
==2315424== by 0x48569A4: ncplane_yx (in /home/dank/src/dankamongmen/notcurses-1.2.3/build/libnotcurses.so.1.2.3)
==2315424== by 0x11190B: fallin_demo (in /home/dank/src/dankamongmen/notcurses-1.2.3/build/notcurses-demo)
==2315424== by 0x10EE7D: main (in /home/dank/src/dankamongmen/notcurses-1.2.3/build/notcurses-demo)
==2315424==
==2315424== HEAP SUMMARY:
==2315424== in use at exit: 20,812,540 bytes in 1,685 blocks
==2315424== total heap usage: 127,843 allocs, 126,158 frees, 1,455,531,300 bytes allocated
==2315424==
==2315424== LEAK SUMMARY:
==2315424== definitely lost: 15,812,416 bytes in 124 blocks
==2315424== indirectly lost: 0 bytes in 0 blocks
==2315424== possibly lost: 634,992 bytes in 4 blocks
==2315424== still reachable: 4,365,132 bytes in 1,557 blocks
==2315424== suppressed: 0 bytes in 0 blocks
==2315424== Rerun with --leak-check=full to see details of leaked memory
==2315424== |
Term: 104x52 alacritty (alacritty terminal emulator)
==2321024== Invalid read of size 4
==2321024== at 0x485CA5B: ncplane_yx (notcurses.c:1840)
==2321024== by 0x113687: drop_bricks (fallin.c:38)
==2321024== by 0x113E03: fallin_demo (fallin.c:193)
==2321024== by 0x110C5D: ext_demos (demo.c:212)
==2321024== by 0x1124DD: main (demo.c:524)
==2321024== Address 0x14 is not stack'd, malloc'd or (recently) free'd |
|
Got it. |
dankamongmen
added a commit
that referenced
this issue
Mar 19, 2020
I noticed while running benchmarks against ranges of widths that we failed on certain widths. I traced this down to a repeatable failure when run with the -c flag at the 104x52 resolution. At this geometry and PRNG seed, we NULLed out the first subwindow of fallin', but didn't recompress the window array. On the next loop we thus used it again, and segfaulted out. This fixes that up. We now pass 100% of the benchmarks, whereas we were failing about 3% before #407.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
For some graphs for the book, I've been running
notcurses-demoin various terminals at geometries of Nx52, where 80 <= N < 192. I run all of these N for each terminal, three times each. The numbers are almost shockingly reproducible (see attached). Anyway, no matter the emulator, we write an empty JSON file about 1--2% of the time. I'm not sure whether we're crashing or what, but this needs be investigated and resolved.The text was updated successfully, but these errors were encountered: