[Bug gdb/30847] gdbtypes.c:3355: internal-error causes gdb to abort when setting breakpoint

["tromey at sourceware dot org" <sourceware-bugzilla@sourceware.org>]

https://sourceware.org/bugzilla/show_bug.cgi?id=30847

--- Comment #3 from Tom Tromey <tromey at sourceware dot org> ---
(In reply to David Brumley from comment #2)
> Thanks for the reply!

No problem, & thank you too.

> This is an old executable and was trying to run as-is.  I have a very weird
> use case. Was demo'ing exploitation (I'm a prof at CMU; demo'ing
> CVE-2020-13995), and was trying to do this on the binary from the vendor.  A
> little more "authentic" that way. In the grand scheme of things this is odd,
> and reported because gdb said to and I was curious if it could be used for
> anti-debugging. Totally fair to close this issue since I can't see this
> happening in any normal dev scenario.

Well, I'm curious to know more about your situation.
We're debating whether to remove stabs support entirely from gdb.
I'm pro-deletion, since it has been obsolete since "forever" (20 years)
and since nobody knows or works on the stabs code -- as you can see
this has resulted in bit-rot.

However, there are others asking that it be kept alive.

One question I have is why you tried a newer gdb rather than an
older one.  But maybe you already answered -- I guess that you
have a new machine but an old executable.

> * I edited the binary to run (and it runs fine) by changing the errno symbol
> to point to stdin. 

Wow.

> I thought the symbol editing might be the source of the problem.  I
> recompiled gdb on my debian system with symbols, and here is the symbol bt
> in case it's useful.  I'm not seeing anything specific to stabs, but I'm
> also a total newb here and don't know anything really.
> 

> #9  0x00005652ee8e562f in define_symbol (valu=0x0, 
>     string=0x5652f01de7d3 "complex double:t(0,17)=r(0,17);16;0;", desc=0,
> type=128, 
>     objfile=0x5652f01a2c40) at stabsread.c:1205

Any stack trace through stabsread.c means you do have stabs.
You can double-check with "readelf -WS" and look for the stabstr section.

You can check your build logs to see what debug flags were used.

-- 
You are receiving this mail because:
You are on the CC list for the bug.