Re: workers/40626 (commit 6c476c22) causes multiple test failures

[Sebastian Gniazdowski <psprint3@xxxxxxxxxxxx>]

On Mon, Feb 27, 2017, at 08:42 AM, Bart Schaefer wrote:
> On Feb 27,  1:41am, Sebastian Gniazdowski wrote:
> } On 26 February 2017 at 20:42, Bart Schaefer <schaefer@xxxxxxxxxxxxxxxx>
> } wrote:
> } > I suspect this is what comes of some attempt to optimize assignments.
> } 
> } It's impossible, hashes weren't optimized.
> 
> Sorry, didn't mean to imply it was the *recent* attempts.

Ah realized that slightly after a while when thought about special
hashes and how internal "ht" is.

> } BTW options+=( ), etc. still segfaults.
> 
> I am not able to reproduce that.  The functions+= test in V06paramter
> passes for me.  Also BEFORE my patch valgrind would report errors:
> torch% options+=()
> torch% options+=()
> ==2255== Invalid read of size 4

I cannot reproduce too, I might undergo some chaos debugging, sorry.
Tried Daniel's patch, HEAD, both with and without my zrealloc hanging
around and nothing. However I can crash zsh-5.2-dev-2 that I'm normally
using. Have 3 cores in /cores from morning, their bts:

 * frame #0: 0x000000010bc02aeb zsh`createparam + 75
    frame #1: 0x000000010bc069ca zsh`arrhashsetfn + 394
    frame #2: 0x000000010bc071e7 zsh`assignaparam + 1031
    frame #3: 0x000000010bbcd14c zsh`addvars + 924
    frame #4: 0x000000010bbcf8f5 zsh`execcmd_exec + 5109

 * frame #0: 0x000000010a987d9b zsh`getarg + 2235
    frame #1: 0x000000010a98720e zsh`getindex + 286
    frame #2: 0x000000010a98954b zsh`fetchvalue + 715
    frame #3: 0x000000010a9ae897 zsh`paramsubst + 5735
    frame #4: 0x000000010a9aba81 zsh`stringsubst + 1777
    frame #5: 0x000000010a9aafa4 zsh`prefork + 180

 * frame #0: 0x00007fc86cae5530
    frame #1: 0x000000010a63020e zsh`getindex + 286
    frame #2: 0x000000010a63254b zsh`fetchvalue + 715
    frame #3: 0x000000010a657897 zsh`paramsubst + 5735
    frame #4: 0x000000010a654a81 zsh`stringsubst + 1777
    frame #5: 0x000000010a653fa4 zsh`prefork + 180

I could think it's zsh-5.2-dev-2 after all, not 5.3.1, because 5.2-dev-2
yielded now:

 * frame #0: 0x00007fdf704459f0
    frame #1: 0x000000010ed1dfde zsh-5.2-dev-2`getindex + 286
    frame #2: 0x000000010ed2031b zsh-5.2-dev-2`fetchvalue + 715
    frame #3: 0x000000010ed45ce2
    zsh-5.2-dev-2`___lldb_unnamed_function208$$zsh-5.2-dev-2 + 5442
    frame #4: 0x000000010ed43011
    zsh-5.2-dev-2`___lldb_unnamed_function204$$zsh-5.2-dev-2 + 1777
    frame #5: 0x000000010ed42534 zsh-5.2-dev-2`prefork + 180
    frame #6: 0x000000010ed432b2 zsh-5.2-dev-2`singsub + 66

However lldb "image list" from the 3 first cores shows:
[ 53] CD26917F-349B-359F-A44C-E9021703D9B5 0x000000010bf4b000
/usr/local/lib/zsh/5.3.1-dev-0/zsh/sched.so
...
[ 53] CD26917F-349B-359F-A44C-E9021703D9B5 0x000000010accf000
/usr/local/lib/zsh/5.3.1-dev-0/zsh/sched.so
...
[ 53] CD26917F-349B-359F-A44C-E9021703D9B5 0x000000010a978000
/usr/local/lib/zsh/5.3.1-dev-0/zsh/sched.so

But if valgrind doesn't report, I've must have some pre-Daniel patch
checkout, sorry again.

PS. Ah, I probably didn't do make install. Ran Src/zsh, but modules were
from /usr/local.

-- 
Sebastian Gniazdowski
psprint3@xxxxxxxxxxxx