RFR (S): 8140424: don't prefix developer and notproduct flag variables with CONST_ in product builds
Christian Thalinger
christian.thalinger at oracle.com
Tue Nov 10 20:44:48 UTC 2015
> On Nov 9, 2015, at 10:41 AM, Lindenmaier, Goetz <goetz.lindenmaier at sap.com> wrote:
>
> Hi Christian,
>
> The function implementation is protected by #ifdef ASSERT, and the call is protected by #ifndef PRODUCT.
> Without this NOT_DEBUG_RETURN it would not compile in the optimized build.
> Another possibility to fix this is checking for #ifndef PRODUCT in memset_with_concurrent_readers.cpp.
> Looking at it now, it must have been broken before.
Thanks for verifying. I will push this as is.
>
> Best regards,
> Goetz.
>
>
>
> From: Christian Thalinger [mailto:christian.thalinger at oracle.com]
> Sent: Monday, November 09, 2015 6:06 PM
> To: Lindenmaier, Goetz <goetz.lindenmaier at sap.com>
> Cc: hotspot-dev at openjdk.java.net; Stefan Karlsson <stefan.karlsson at oracle.com>
> Subject: Re: RFR (S): 8140424: don't prefix developer and notproduct flag variables with CONST_ in product builds
>
> I did verify that our compilers are in fact eliminating all code and Stefan Karlsson made sure some important GC functions still produce the same code.
>
> On Nov 8, 2015, at 9:40 PM, Lindenmaier, Goetz <goetz.lindenmaier at sap.com<mailto:goetz.lindenmaier at sap.com>> wrote:
>
> Hi Christian,
>
> thanks for incorporating these cleanups.
> The change still looks good ☺.
>
> Best regards,
> Goetz.
>
> From: Christian Thalinger [mailto:christian.thalinger at oracle.com]
> Sent: Freitag, 6. November 2015 18:13
> To: Lindenmaier, Goetz
> Cc: hotspot-dev at openjdk.java.net<mailto:hotspot-dev at openjdk.java.net>
> Subject: Re: RFR (S): 8140424: don't prefix developer and notproduct flag variables with CONST_ in product builds
>
>
> On Nov 6, 2015, at 1:38 AM, Lindenmaier, Goetz <goetz.lindenmaier at sap.com<mailto:goetz.lindenmaier at sap.com>> wrote:
>
> Hi Christian,
>
> thanks for doing this change. It looks good.
>
> It's good that now notproduct flags are also defined in the product build as constants,
> as it's already the case with debug flags.
> This allows more cleanups in the style you did to compilerDirectives.hpp.
>
> I had a look at these cleanups, quite a row you can find in this patch:
> http://cr.openjdk.java.net/~goetz/webrevs/8140424-notProd/not_product_cleanups.patch
>
> That is a very nice cleanup.
>
>
>
>
> Does it make sense to commit these along with your change?
>
> It does. I’ve updated my webrev with your changes:
>
> http://cr.openjdk.java.net/~twisti/8140424/webrev/
>
>
>
>
> Best regards,
> Goetz.
>
>
>
>
>
> -----Original Message-----
> From: hotspot-dev [mailto:hotspot-dev-bounces at openjdk.java.net] On
> Behalf Of Christian Thalinger
> Sent: Freitag, 6. November 2015 07:23
> To: hotspot-dev at openjdk.java.net<mailto:hotspot-dev at openjdk.java.net>
> Subject: RFR (S): 8140424: don't prefix developer and notproduct flag
> variables with CONST_ in product builds
>
> https://bugs.openjdk.java.net/browse/JDK-8140424
> <https://bugs.openjdk.java.net/browse/JDK-8140424>
> http://cr.openjdk.java.net/~twisti/8140424/webrev/
> <http://cr.openjdk.java.net/~twisti/8140424/webrev/>
>
> When working on JDK-8024545 I added code that prefixes developer and
> notproduct flags with CONST_. I think I did this to be able to access the
> address of these variables for the flag table. But this is not necessary because
> these flag cannot be written anyway.
>
> I am proposing to remove the CONST_ prefix and define the variables as
> const.
>
More information about the hotspot-dev
mailing list