Probable bug

[Takashi Yano]

On Mon, 31 Jul 2023 18:41:06 +0900
Takashi Yano wrote:
> On Sun, 30 Jul 2023 20:37:35 +0900
> Takashi Yano wrote:
> > On Sun, 30 Jul 2023 07:29:10 -0400
> > Eliot Moss wrote:
> > > On 7/30/2023 5:38 AM, natan_b--- via Cygwin wrote:
> > > > Hi Guys
> > > > 
> > > > very short.
> > > > 
> > > > prog.c
> > > > 
> > > > #include <stdio.h>
> > > >   
> > > > int main()
> > > > {
> > > > float a=1.283;
> > > > while(1)
> > > > printf( "%f", a );
> > > > }
> > > >   
> > > > run with
> > > > $ ./prog.exe >/dev/null
> > > >   
> > > > in windows monitor process the process increase it's memory it arrive to many  Gb.
> > > > It's not a machine problem, other PC have same problem.
> > > >   
> > > > Same program in wsl and MSYS2 works well!
> > > 
> > > This probably has to do with output buffering, and may happen even without
> > > the >/dev/null since there are no line ends in the output.  It may work with
> > > stdbuf -o0 (as in: stdbuf -o0 ./prog.exe >/dev/null) but may cause the program
> > > to run more slowly (each character is sent to the device, when then immediately
> > > discards it).  It would seem you're hoping for the internal libraries to
> > > recognize the case of writing to /dev/null ...
> > 
> > I also suspected that, however, it was not correct.
> > 
> > while (1) sprintf(buf, "%f\n", a);
> > 
> > has the same problem. :-(
> 
> I looked into this problem, and found the cause.
> This is the serious bug in the commit:
> 
> commit a4705d387f7874f9a5de3ee67861773a17691643
> Author: Takashi Yano <takashi.yano@nifty.ne.jp>
> Date:   Wed Dec 1 06:11:32 2021 +0900
> 
>     ldtoa: Import gdtoa from OpenBSD.
> 
>     - This patch uses gdtoa imported from OpenBSD if newlib configure
>       option "--enable-newlib-use-gdtoa=no" is NOT specified.  gdtoa
>       provides more accurate output and faster conversion than legacy
>       ldtoa, while it requires more heap memory.
> 
> Please let me consider how to fix that.

I have just posted a patch for this issue to newlib mailing list.

-- 
Takashi Yano <takashi.yano@nifty.ne.jp>