bug-gawk
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [bug-gawk] Increased RAM memory usage in gawk 5.0.1 compared to gawk


From: arnold
Subject: Re: [bug-gawk] Increased RAM memory usage in gawk 5.0.1 compared to gawk 4.1.4
Date: Fri, 16 Aug 2019 02:58:52 -0600
User-agent: Heirloom mailx 12.5 7/5/10

Glad it's working. I have pushed a cleaner fix to git for the
next release.

Thanks for the report.

Arnold

Finn Magnusson <address@hidden> wrote:

>  Thank you ! I could now apply the patch and it fixed the memory leak issue !
> Thank you for the great support :-)
> Best regards
> Finn
>
>     On Thursday, August 15, 2019, 8:30:36 AM GMT+2, address@hidden 
> <address@hidden> wrote:  
>  
>  Using
>
>     patch -p1 < the-patch-file
>
> should have worked. If not, you can simply manually edit builtin.c
> and add the bracketing C preprocessor control lines.
>
> Thanks,
>
> Arnold
>
> Finn Magnusson <address@hidden> wrote:
>
> >  Hi !
> > Thank you for the patch and for the tip about the pipe , will correct our 
> > code accordingly.Regarding the patch, I tried to run it on gawk-5.0.1 but 
> > it did not seem to run, see below. 
> > Did I miss something or should I run it on a different gawk version?
> > seliius01071 [20:16] [eanzmagn/gawk501/gawk-5.0.1] -> patch -p1 -i 
> > arnold.patch 
> > patching file builtin.c
> > Hunk #1 FAILED at 1647.
> > 1 out of 1 hunk FAILED -- saving rejects to file builtin.c.rej
> > seliius01071 [20:16] [eanzmagn/gawk501/gawk-5.0.1] -> cat arnold.patch 
> > diff --git a/builtin.c b/builtin.c
> > index 1c205aa..0ddfb32 100644
> > --- a/builtin.c
> > +++ b/builtin.c
> > @@ -1647,8 +1647,10 @@ mpf1:
> >     }
> >     bchunk(s0, s1 - s0);
> >     olen_final = obufout - obuf;
> > +#if 0
> >     if (ofre > 0)
> >         erealloc(obuf, char *, olen_final + 1, "format_tree");
> > +#endif
> >     r = make_str_node(obuf, olen_final, ALREADY_MALLOCED);
> >     obuf = NULL;
> >
> >  seliius01071 [20:17] [eanzmagn/gawk501/gawk-5.0.1] -> cat builtin.c.rej 
> > --- builtin.c
> > +++ builtin.c
> > @@ -1647,8 +1647,10 @@
> >     }
> >     bchunk(s0, s1 - s0);
> >     olen_final = obufout - obuf;
> > +#if 0
> >     if (ofre > 0)
> >         erealloc(obuf, char *, olen_final + 1, "format_tree");
> > +#endif
> >     r = make_str_node(obuf, olen_final, ALREADY_MALLOCED);
> >     obuf = NULL;
> > seliius01071 [20:18] [eanzmagn/gawk501/gawk-5.0.1] -> 
> >
> >  
> >
> >
> >
> > Thank youFinn
> >
> >
> >    On Tuesday, August 13, 2019, 8:08:29 PM GMT+2, address@hidden 
> ><address@hidden> wrote:  
>
> >  Hi.
> >
> > Finn Magnusson via bug-gawk <address@hidden> wrote:
> >
> > > Dear gawk developers
> > >
> > > Our team recently started using gawk 5.0.1 and found that the RAM memory
> > > usage seems to have increased compared to gawk version 4.1.4, which we
> > > were using before.
> >
> > The patch below seems to fix the problem.  I am not sure why.
> > I will continue to investigate, but in the meantime I think this will
> > get you going. Please let me know.
> >
> > By the way, your test program uses `|&' which opens a two-way pipe
> > to the subprocess. However, all you're doing is reading from it, so
> > it's enough to just use `|'.
> >
> > Thanks,
> >
> > Arnold
> > ------------------------------------------------------------
> > diff --git a/builtin.c b/builtin.c
> > index 1c205aa..0ddfb32 100644
> > --- a/builtin.c
> > +++ b/builtin.c
> > @@ -1647,8 +1647,10 @@ mpf1:
> >      }
> >      bchunk(s0, s1 - s0);
> >      olen_final = obufout - obuf;
> > +#if 0
> >      if (ofre > 0)
> >          erealloc(obuf, char *, olen_final + 1, "format_tree");
> > +#endif
> >      r = make_str_node(obuf, olen_final, ALREADY_MALLOCED);
> >      obuf = NULL;
> >  out:
>
>   



reply via email to

[Prev in Thread] Current Thread [Next in Thread]