Minutes from 17 August meeting

[Jim Dehnert]

Hi, all,

Since our attendance was, ah, weak, I'll summarize what we did.
And there's homework below, so please read...  Numbers are from the
agenda.

 2) C-2:  We accepted the current version of constructor priority,
    with the additional change (I missed it last time) of the pragma
    name from init_priority to just priority.  I have verified since
    that this won't conflict with anything from OpenMP or pthreads.

 3) C-18:  We accepted the proposal.  That is, in all cases where a
    result class is to be returned in a buffer supplied by the caller,
    whether for long classes (>256 bits) or because of non-trivial copy
    constructors or destructors, the buffer address is passed in r8 on
    IA-64, as specified by the psABI for long structs in C.

 9) H-1:  We will name the C++ runtime library libcxa.so .

Now for the homework:

Our next meeting is scheduled for 31 August, just before Labor Day
weekend.  I would like to go over any issues people have found with the
exception handling draft at the next meeting.  So, if you (a) are
concerned with that subject, and (b) can't come 31 August or can't look
at it carefully by then, please let me know ASAP, and I'll consider
cancelling that meeting.

The following scheduled meeting is 14 September, but I'll be out of
town and intend to cancel it.  So the next one after 31 August will be
28 September.

 1) Alex:  Does the mangling grammar look OK now?

 4) Mark:  We need a specific list of changes for C-19.

 5) Christophe:  Does the proposal at the end of G-4 (open issues page)
    look like it solves your concerns?    Would anyone else have
    problems with it as a resolution of the one-time init interface?
    Note the tradeoff:  It would increase the guard variable size from
    8 to 40 bytes, but would likely eliminate a bunch of instructions
    to gather that data for the destructor registration call.  (But it
    would be a pure loss for no-destructor objects.  So perhaps we
    should modify it to eliminate the extra data for those, and pass a
    parameter to indicate that to the release routine?)

 6) Everyone:  Go over the exception handling material.

 7) Christophe:  I think issue F-5 is dead unless you have something
    specific in mind (mangling ILP32 vs. LP64), and probably even then.
    Do you.

 8) Christophe:  Similarly G-1 on command line options.

I'm on vacation next week, so I won't see or respond to anything
between the 19th and 28th.  I'll try to get updated documents out by
tomorrow.

Jim

-		Jim Dehnert  x3-4272