Re: Segfault in texinfo

[Gavin Smith]

On Sun, Jan 22, 2023 at 06:57:22AM +0000, Sam James wrote:
> 
> 
> > On 21 Jan 2023, at 21:35, jake roggenbuck <roggenbuckjake@gmail.com> wrote:
> > 
> > To whom it may concern,
> > 
> > I have found a crash and segfault in texinfo (info).
> > 
> > Here is how you reproduce the segfault.
> > 
> > ```
> > cat /dev/urandom | info
> > ```
> > 
> 
> Please capture some exact input to trigger the segfault (pipe to a file 
> first, then test it after by piping into info).
> 

Yes, this has been reported in the past and it is an interesting way to
find bugs, but this does not immediately help us to fix them.

Working on this is a low priority for me as the random input may be triggering
rarely used code or in ways that are unlikely to occur in practice.

At present, it would be quite a lot of work to try to track these bugs down.
It would be good if we could make this easier by working out a routine
way of debugging this.

One way may be to use tee:

  cat /dev/urandom  | tee dump | info

Then progressively test smaller subsets of "dump", taking bytes off the
end first, then off the beginning:

  cat dump.subset | info

Once dump is at a minimal size (say less than 50 bytes) then it's important
to know whether particular Info files are being loaded, and the failure may
be dependent on which Info files are on your system and the contents of
your dir files.

Valgrind is a good program for identifying the exact point in a program's
execution where a memory error first occurs.

As you can tell, there are a lot of details to work out here.