Subject: Jan 12 Notes, next meeting
From: Hans Boehm (boehm_at_[hidden])
Date: 2021-01-12 19:53:39
We discussed D1875R2, the work-in-progress revision of P1875.
There was agreement that
1) We need to be more explicit about advocating for the inclusion of memory
2) We need to allow out-of-memory (bad_alloc) exceptions to be thrown
inside transactions, so long as they don't exit the transaction.
Michael Scott expressed concerns that declaring bad_alloc thrown out of a
transaction to be "undefined behavior" would encourage implementations to
not properly report such failures. On the other hand, we can't easily
insist that terminate() be called, since the handler is replaceable, and
the transaction state will presumably have been reverted, making it hard to
define what that means.
Mike Spear and Hans Boehm discussed the phrasing of the last paragraph in
6.0 and came up with a direction.
Consult with Jens as to whether "undefined behavior" for escaping
exceptions is a real issue. Is there a better way to handle this?
Hans to thoroughly read the paper, clear comments where possible, and edit
to take more of a stand on allocation.
Feb 9, 2021 8:00 PST, 11:00 EST
(âªUSâ¬) âª+1 208-925-0196â¬
PIN: âª255 542#â¬
(or ask for other phone numbers)
SG5 list run by firstname.lastname@example.org
Older Archives on Google Groups