Core dumped on an undefined variable with a new PC definition

[docbacardi]

Core dumped on an undefined variable with a new PC definition

[thecky]

Hi, I checked out the commit ba43fe1 - which is the last commit in 2014 (= oldversion), and I can't reproduce your issue.
I used your code snippet from above and got this output:
Some vars still undefined. Dump not yet supported, sorry! testfile.asm:6: error:variable undefined: show

Even with the latest commit (bc34eb3), I get the desired error message.

[pawelkrol]

OK, so I still get a segmentation fault on ba43fe1 just as described:

$ git reset --hard ba43fe1
HEAD is now at ba43fe1 Code cleanups. Use fixed sizes for all types.
$ make
$ ./dreamass program.src 
Some vars still undefined. Dump not yet supported, sorry!
dreamass: segment.c:430: segment_enterSegElem: Assertion `false' failed.
Aborted (core dumped)

Interestingly it doesn't show up under valgrind:

$ valgrind --leak-check=full dreamass program.src 
Some vars still undefined. Dump not yet supported, sorry!
WARNING: PC still phased at end of segment.
program.src:6: error:variable undefined: show

However after fetching just the very next commit 98c93ba everything works as expected:

$ git merge $(git rev-list master..origin/master | tail -n1)
Updating ba43fe1..98c93ba
Fast-forward
 segment.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
$ make
$ ./dreamass program.src 
Some vars still undefined. Dump not yet supported, sorry!
WARNING: PC still phased at end of segment.
program.src:6: error:variable undefined: show

Judging from the contents of a core dump, it looks like the issue has been indeed fixed by 98c93ba:

$ ulimit -c unlimited
$ ./dreamass program.src
$ coredumpctl list
TIME                            PID   UID   GID SIG COREFILE  EXE
Wed 2021-04-14 18:05:25 CEST  19072  1000  1000   6 present   /data/Tools/dreamass/dreamass
$ coredumpctl info 19072
        Signal: 6 (ABRT)
       Message: Process 19072 (dreamass) of user 1000 dumped core.
                
                Stack trace of thread 19072:
                #0  0x00007f67c842d9d5 raise (libc.so.6 + 0x3d9d5)
                #1  0x00007f67c84168a4 abort (libc.so.6 + 0x268a4)
                #2  0x00007f67c8416789 __assert_fail_base.cold (libc.so.6 + 0x26789)
                #3  0x00007f67c8426026 __assert_fail (libc.so.6 + 0x36026)
                #4  0x0000000000409daf segment_enterSegElem (dreamass + 0x9daf)
                #5  0x0000000000406852 nextvalid (dreamass + 0x6852)
                #6  0x0000000000406b0b pp_get (dreamass + 0x6b0b)
                #7  0x0000000000405fdd pass_showUndefs (dreamass + 0x5fdd)
                #8  0x0000000000402db2 baekwha (dreamass + 0x2db2)
                #9  0x00000000004013d2 main (dreamass + 0x13d2)
                #10 0x00007f67c84181e2 __libc_start_main (libc.so.6 + 0x281e2)
                #11 0x000000000040141e _start (dreamass + 0x141e)

No clue why you couldn't reproduce the problem. I'm glad it's been fixed though.

[thecky]

Yep, it seems fixed in latest commits - and your're right, the bug is still in ba43fe1.
My mistake was: I used my version under /usr/bin and not the version from git (forgotten the ./) and the version under /usr/bin is from 2014 - but without the bug (it's not a version from Github).