Re: [PATCH] Do not write negative PID or TID in remote protocol

[Tom Tromey <tromey@adacore.com>]

Mark> It looks like there is still something wrong on 32bit arches.
Mark> At least after this patch the i386 debian buildbot fails:
Mark> https://builder.sourceware.org/buildbot/#/builders/105/builds/12989

I'm sorry about that.  This patch turned out to be kind of bad.

I built a -m32 gdb to reproduce this and I'll send the fix shortly.

>> +      hex = hex_or_minus_one (p, &pp, for_remote);
>> +      if (pp == p)
>> +	error (_("invalid remote ptid: %s"), buf);
>> +
>> +      lwp = (ptid_t::lwp_type) hex;
>> +      if (hex != ((unsigned_lwp_type) lwp))
>> +	error (_("invalid remote ptid: %s"), buf);

Mark> Are the types/castes here correct? hex is an ULONGEST that is being
Mark> cast to an ptid_t::lwp_type and then an unsigned_lwp_type. Then this
Mark> sanity check seems to fail (on i386 at least). I think this is because
Mark> ULONGEST on i386 is unsigned long long, while unsigned_lwp_type is
Mark> unsigned long.

Yes, I agree; and I think the fix is to special-case the "-1" value
here.

Tom