Re: segmentation violation in debug mode when watching for an index of an array and rerunning

[Pascal Maugis]

Hi Arnold,

sorry for not having tested that behavior with the top--newest release 
as I should have, but that's cool from you to have explained the 
progress.Having no core dump anymore is less worrying and a kind of 
relief :)

I understand the difficulty of co-developping, while we rely on a few 
guys that might leave anytime. If I were consistent as a user, I would 
enter the debbuger code and make the modification myself. So I'm not 
going to throw stones at anybody.

Anyway, you've done a great job. I can make do with that small bug. 
rerunning is not so annoying, and watch points remain pretty useful and 
manageable that way.

Cheers,

Pascal.

Le 17/09/2022 à 21:19, arnold@skeeve.com a écrit :
> Hello.
>
> Thanks for the report.
>
> Please try the latest release, 5.2.0, made about 2 weeks ago. Here
> is what I get:
>
>         $ cat a.awk
>         a[1] = 2
>
>         $ ./gawk --debug -f a.awk
>         gawk> b 1
>         Breakpoint 1 set at file `a.awk', line 1
>         gawk> run
>         Starting program:
>         
>         Stopping in Rule ...
>         Breakpoint 1, main() at `a.awk':1
>         1       a[1] = 2
>         gawk> watch a[1]
>         Watchpoint 2: a["1"]
>         gawk> run
>         Program already running. Restart from beginning (y/n)? y
>         Restarting ...
>         Starting program:
>         Program exited normally with exit value: 0
>
> Now, exiting right away isn't lovely, but at least it's not a core dump.
> (I will explore it further.)
>
> In general, watchpoints are pretty fragile.  The debugger was written
> by another developer who abandoned the project a long time ago, and is
> perhaps the part of gawk I am least familiar with, unfortunately.
>
> Arnold
>
> Pascal Maugis<pascal.maugis@lsce.ipsl.fr>  wrote:
>
> > In debug mode, watching for operation on the index of an array is
> > possible (et very usefull). However, when rerunning the program while
> > the watch is still active leads to segmentation violation :
> >
> >      zsh> echo "a[1]=2" >! a.awk
> >      zsh> gawk -D -f ./a.awk
> >      *gawk> b *1
> >       ?????? Point d'arret 1 defini dans le fichier ????./a.awk???? ligne 1
> >      *gawk> r*
> >       ?????? Demarrage du programme??:
> >       ?????? Arr??t dans Rule...
> >       ?????? Breakpoint 1, main() ?? ????./a.awk????:1
> >       ?????? 1???????????? a[1]
> >      gawk> n
> >      gawk> w a[1]
> >      Watchpoint 2: a["1"]
> >      gawk> r
> >       ?????? Programme en cours. Reprendre depuis le debut (o/n)??? o
> >      *?????? gawk: fatal??: erreur interne??: erreur de segmentation*
> >       ?????? [2]?????? 1001153 abort (core dumped)?? gawk -D -f ./a.awk
> >
> > Obviously, after rerun, 'watch' points to an adress which is not
> > allocated anymore (which is logic). I guess watchpoints on array indexes
> > should be disabled on rerun, or reenabled only when the array is defined
> > again as such. Watching a whole array ('watch a') works fine.
> >
> > Cheers,
> >
> > Pascal.
> >
> > Le 13/09/2022 ?? 12:42, Pascal Maugis a ??crit??:
> > > Hi, as promised here is the strange, unexpected behavior of ++ that
> > > looks like an error to me, because in "a ++b", a is incremented
> > > instead of b. "a b++" and "a++ b" produce the correct results however :
> > >
> > >      touch a.awk
> > >      gawk -D -f ./a.awk
> > >      eval "a=1 ; b=2"
> > >      eval "print a ++b"???? ; output = 12 : (variables will be
> > >      incremented after the operation, ok with me as documented)
> > >      eval "print a b"???????? ; output = 22 : a has been incremented
> > >      instead of b
> > >      eval "print a++b"?????? ; output = 22 :
> > >      eval "print a b"???????? ; output = 32 : a has been incremented
> > >
> > > The operation seems to be interpreted as "a++b", the space been
> > > uninterpreted as a separator
> > >
> > > Thanks for your insight,
> > >
> > > Cheers,
> > >
> > > Pascal.
> > >
> > > --
> > > ------------------------------------------------------------------------
> > >
> > > *Pascal MAUGIS*
> > > Mod??lisation hydrologique et changement climatique, PhD
> > >
> > > LSCE
> > >
> > > Laboratoire des Sciences du Climat et de l'Environnement
> > > <http://www.lsce.ipsl.fr/>
> > >
> > > Unit?? Mixte de Recherche CEA-CNRS-UVSQ (UMR 8212), IPSL, Paris-Saclay
> > > Equipe MOSAIC : MOdelisation des SurfAces et Interfaces Continentales
> > > B??t. 714, pi??ce 2115
> > > C.E. Saclay, Orme des Merisiers, 91191 Gif-sur-Yvette Cedex, FRANCE
> > >
> > > e-mail :pascal.maugis@lsce.ipsl.fr
> > > Page perso : FR<http://www.lsce.ipsl.fr/Pisp/pascal.maugis/>  - EN
> > > <http://www.lsce.ipsl.fr/en/Phocea/Pisp/index.php?nom=pascal.maugis>
> > > Tel./Fax : +33 1 69 08 75 24 / +33 1 69 08 77 16
> > >
> > > ------------------------------------------------------------------------
> > --
> > ------------------------------------------------------------------------
> >
> > *Pascal MAUGIS*
> > Mod??lisation hydrologique et changement climatique, PhD
> >
> > LSCE
> >
> > Laboratoire des Sciences du Climat et de l'Environnement
> > <http://www.lsce.ipsl.fr/>
> >
> > Unit?? Mixte de Recherche CEA-CNRS-UVSQ (UMR 8212), IPSL, Paris-Saclay
> > Equipe MOSAIC : MOdelisation des SurfAces et Interfaces Continentales
> > B??t. 714, pi??ce 2115
> > C.E. Saclay, Orme des Merisiers, 91191 Gif-sur-Yvette Cedex, FRANCE
> >
> > e-mail :pascal.maugis@lsce.ipsl.fr
> > Page perso : FR<http://www.lsce.ipsl.fr/Pisp/pascal.maugis/>  - EN
> > <http://www.lsce.ipsl.fr/en/Phocea/Pisp/index.php?nom=pascal.maugis>
> > Tel./Fax : +33 1 69 08 75 24 / +33 1 69 08 77 16
> >
> > ------------------------------------------------------------------------
--
------------------------------------------------------------------------

*Pascal MAUGIS*
Modélisation hydrologique et changement climatique, PhD

LSCE

Laboratoire des Sciences du Climat et de l'Environnement 
<http://www.lsce.ipsl.fr/>

Unité Mixte de Recherche CEA-CNRS-UVSQ (UMR 8212), IPSL, Paris-Saclay
Equipe MOSAIC : MOdelisation des SurfAces et Interfaces Continentales
Bât. 714, pièce 2115
C.E. Saclay, Orme des Merisiers, 91191 Gif-sur-Yvette Cedex, FRANCE

e-mail : pascal.maugis@lsce.ipsl.fr
Page perso : FR <http://www.lsce.ipsl.fr/Pisp/pascal.maugis/> - EN 
<http://www.lsce.ipsl.fr/en/Phocea/Pisp/index.php?nom=pascal.maugis>
Tel./Fax : +33 1 69 08 75 24 / +33 1 69 08 77 16

------------------------------------------------------------------------