Hi Thomas, the change looks good, but I'd prefer if you leave the initial comment in place which also mentions "-qminimaltoc" as a way of resolving TOC overflow errors.
I actually don't remember exactly, but I think "-qminimaltoc" works by creating distinct TOCs for each compilation unit. That comes with an performance impact, but "-qpic=large" / "-bbigtoc" can have an performance impact as well. As I said, for the slowdebug build your changes are fine. I'd just like to keep the reference to "-qminimaltoc" for the case where we have to re-evaluate the different solutions for the product build. Thank you and best regards, Volker On Mon, Jul 17, 2017 at 3:19 PM, Langer, Christoph <christoph.lan...@sap.com> wrote: > Hi Thomas, > > the fix looks ok to me. > > I’m copying build-dev because of the changes in generated-configure.sh. Don’t > know if this can just be pushed from extern or if it needs some special > handling from Oracle folks or other things to take care of? > > Best regards > Christoph > > From: ppc-aix-port-dev [mailto:ppc-aix-port-dev-boun...@openjdk.java.net] On > Behalf Of Thomas Stüfe > Sent: Montag, 17. Juli 2017 12:55 > To: ppc-aix-port-...@openjdk.java.net > Subject: RFR(xs): 8184344: [aix] libjvm.so TOC overflow for slowdebug > > Hi all, > > may I please have a review for the following fix: > > webrev: > http://cr.openjdk.java.net/~stuefe/webrevs/8184344-TOC-overflow/webrev.00/webrev/ > Issue: https://bugs.openjdk.java.net/browse/JDK-8184344 > > Basically, the TOC on AIX overflows on slowdebug builds. I am not sure yet > which change caused that - I suspect one of the recent > template-metaprogramming changes but have not investigated yet. > > Thanks, Thomas >
