bug-texinfo
[Top][All Lists]
Advanced

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

Re: address@hidden: Bug#79091: info: "info --vi-keys libc:exit" causes s


From: Eli Zaretskii
Subject: Re: address@hidden: Bug#79091: info: "info --vi-keys libc:exit" causes segfault when you hit 'h']
Date: Tue, 12 Dec 2000 11:17:42 +0200 (IST)

On Tue, 12 Dec 2000, Adam Olsen wrote:

> it crashes when I have "info --vi-keys" plus some random (non-argument)
> text.  I havn't tried with any valid arguments.  It doesn't need a colon to
> crash, but requires I hit 'h' twice if I don't have one.  I'm not sure if
> it's only with invalid entries, since I don't know any valid ones to try it
> with. :)

Thanks for the info and the backtraces, I will try to find this sucker.

> then a bit later in ddd I got this, although the first couple times I tried
> it didn't crash.
> 
> (gdb) bt
> #0  0x400cd3d8 in free () from /lib/libc.so.6
> #1  0x400cd2c3 in free () from /lib/libc.so.6
> #2  0x804d90e in gc_pointers () at gc.c:91

Sigh.  This definitely looks like I was right in my guess: some code 
somewhere is corrupting the memory chain, and causes malloc/free to crash 
in random places...

> Looking for patterns in this now.  If I do the same one over a couple times
> it appears to have the same backtrace.

That's what I'd expect: the same command-line argument causes the same 
chain of calls to malloc/free with the same arguments.  So it crashes in 
the same spot.

> Also, if I use "we" as my random
> string I get a page for "Web2c", and it won't crash.

That's a feature.  Info allows you to ``shorthand'' the menu entries: if 
it doesn't find the exact match in the menu (in this case, "we" doesn't 
exist in DIR), it looks for it ``sloppily'', by matching the string you 
type to the beginning of every menu entry.  This allows to find, say, 
"Invoking Foo" when the user only types "Invoking".

> (btw, my bet's on how it handles invalid entries at this point, but I'm not
> sure how that ties in with --vi-keys...)

Me neither.  --vi-keys changes the text printed when you ask for help 
(because the key bindings are different).  Perhaps some of that code is 
bogus...

But --vi-keys could also be not really relevant: it might be that the bug 
is somewhere else, and only turns up with --vi-keys because the 
additional argv[] element changes the memory-allocation pattern in the 
libc startup code.



reply via email to

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