BP status?

[Andreas Jaeger]

>>>>> Greg McGary writes:

Greg> Andreas Jaeger <aj@suse.de> writes:
>> Could you give some details about the current status?

Greg> I'm mostly working on getting glibc changes checked in, which will
Greg> probably take another week or so.  I'll announce when it's ready for
Greg> testing.

Thanks.

Greg> After glibc, I have a big pile of gcc changes to commit.  That might
Greg> take many weeks, since patch approval latency is much longer.
Greg> However, BP gcc is available right away since I have all my changes on
Greg> a branch.
Ok, I'll check it out soonish.

>> Does it make sense to test builds with bounded pointers?

Greg> Not yet.  Testing BP builds will be a good idea as soon as I have all
Greg> glibc changes checked in.

>> On which architectures does it work?

Greg> It mostly works on ix86: gcc bootstraps with BP glibc.  It almost
Greg> works on PowerPC: there are still a few gcc bootstrap problems.
Greg> Targets I will work on next are MIPS and i960.  After that, if I have
Greg> time, and no one else has done them yet, I'll do on IA-64, SPARC and
Greg> Alpha.

>> I guess I need the current gcc from CVS for this - or
>> do I need extra patches for GCC?

Greg> My compiler changes are on a CVS branch: bounded-pointers-branch.

>> I'd appreciate if you could write some documentation about the bounded
>> pointer support when you're finished.  This should be added to the
>> manual and perhaps also to the FAQ.

Greg> I don't see that there's much to write about BPs in the glibc manual,
Greg> just as there's very little to write about the static debug or
Greg> profiling versions of glibc: they implement API-compatible drop-in
Greg> replacements and are thus transparent to the user.  By contrast, the
Greg> debug and profiling versions are also ABI-compatible whereas the BP
Greg> version is not.  IMO, BPs are more properly documented in the gcc
Greg> manual and the glibc manual need only refer the reader to the gcc
Greg> manual.  What do you think?

Do you need to compile glibc with bp support if you want to use it in
user programs?  We should at least explain the --enable-bounded
parameter in the manual (see INSTALL/install.texi).

If everything is explained in gcc, there's no further need to document
it - I'll look later again into it.

Andreas
-- 
 Andreas Jaeger
  SuSE Labs aj@suse.de
   private aj@arthur.inka.de