Re: [bug-gawk] bug related to hash function

[Aharon Robbins]

Hi.

Status:

1. I can reproduce this. This is good.

2. It only takes a few lines of your data. This is also good.

3. Using git bisect, I found the commit at which the behavior broke. Also good.

4. It was a big commit that reworked the array handling, and the person who
originally wrote the code is long gone from the project. This is bad.

I have made progress tracking down the bug and I see at what point
in the awk program that gawk goes wrong. I will pick up debugging from
there later, to try to see what's happening inside gawk. I hope that I
can fix it.

Thanks,

Arnold

> Subject: Re: [bug-gawk] bug related to hash function
> To: Aharon Robbins <address@hidden>
> Cc: address@hidden
> From: Andres Legarra <address@hidden>
> Date: Tue, 3 May 2016 10:58:08 +0200
> 
> Update, I have tried with 4.1.3 from macports, to no success.
> 
> On 02/05/16 20:53, Aharon Robbins wrote:
> > Hello.
> >
> > Thanks for the report. Please send a full program that demonstrates
> > the bug, and any necessary data files as well. I can't attempt to
> > diagnose any such issue if I can't first reproduce it.
> >
> > You might also try with the current release, which is 4.1.3.
> >
> > Also, did you compile gawk as a 32 bit or 64 bit executable?
> >
> > Thanks,
> >
> > Arnold
> >
> >> To: address@hidden
> >> From: Andres Legarra <address@hidden>
> >> Date: Mon, 2 May 2016 17:34:36 +0200
> >> Subject: [bug-gawk] bug related to hash function
> >>
> >> Hi there,
> >>
> >> I have found a weird bug concerning hashes. For a particular element
> >> x="00000779770060", the value of its hash table
> >> posout[x]
> >> that should be 44, becomes empty (null string). The asignment of this 44
> >> works, but then it gets lost.
> >>
> >> If I print it to /dev/stderr every few loops, it is not lost.
> >>
> >> I see this behaviour in my mac running El Capitan and GNU Awk 4.1.0,
> >> API: 1.0
> >> However in our Linux server running GNU Awk 3.1.7 I do not see this problem
> >>
> >> The error is reproducible (in my machine) and I can send files if needed.
> >>
> >> any help is welcome
> >> Andres
> >> -- 
> >> Andres Legarra
> >> +33 561285182
> >> INRA, UMR1388 GenPhySE
> >> CS 52627
> >> 31326 Castanet Tolosan, France
> 
> -- 
> Andres Legarra
> +33 561285182
> INRA, UMR1388 GenPhySE
> CS 52627
> 31326 Castanet Tolosan, France
>