On 24 February 2014 17:02, Daniel Dunbar <daniel at zuster.org> wrote:> Nothing in particular springs to mind. Is it possible that in the context > where this part of the Makefile is running Makefile.config isn't available?This is a standard lnt run, create in the same way I ran for x86_64 and ARM, so I'm not sure what could be wrong.> I'm not sure how this can be completely broken in the fashion you describe > though, as it seems like that would make little-endian fail which is what > most people are testing.It is very weird. I got lost twice trying to follow the make files and autoconf, this is why I ask you, to see if you can think of anything that would need registering when adding new targets... I'll keep trying... Thanks! --renato
I can't think of whats going on, but to debug I would: 0. Verify your test suite checkout is clean and doesn't have any temporary or build products. 1. Use the 'lnt runtest nt' --only-test option for a specific subdir to be able to iterate quickly. 2. Try adding debug statements to the Makefile to dump the value of ENDIAN, e.g., with: $(error "ENDIAN is $ENDIAN"), before and after the include of Makefile.config and Makefile.singlesrc (which includes Makefile.programs). - Daniel On Mon, Feb 24, 2014 at 9:32 AM, Renato Golin <renato.golin at linaro.org>wrote:> On 24 February 2014 17:02, Daniel Dunbar <daniel at zuster.org> wrote: > > Nothing in particular springs to mind. Is it possible that in the context > > where this part of the Makefile is running Makefile.config isn't > available? > > This is a standard lnt run, create in the same way I ran for x86_64 > and ARM, so I'm not sure what could be wrong. > > > > I'm not sure how this can be completely broken in the fashion you > describe > > though, as it seems like that would make little-endian fail which is what > > most people are testing. > > It is very weird. I got lost twice trying to follow the make files and > autoconf, this is why I ask you, to see if you can think of anything > that would need registering when adding new targets... > > I'll keep trying... Thanks! > > --renato >-------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20140224/229a5daf/attachment.html>
On 24 February 2014 18:32, Daniel Dunbar <daniel at zuster.org> wrote:> $(error "ENDIAN is $ENDIAN"), before and after the include of > Makefile.config and Makefile.singlesrc (which includes Makefile.programs).Hi Daniel, Thanks, that did the trick! ;) In TargetConfig.mk.in: #ifdef __LITTLE_ENDIAN__ ENDIAN := little #else ENDIAN := big #endif Seems like it should be __ORDER_LITTLE_ENDIAN__ to be cross-platform. Attaching the patch that adds aarch64 (which is identical to the one applied in LLVM a few days ago) to the config plus change the TargetConfig.mk. Shouldn't cause too much disturbance. ;) cheers, --renato -------------- next part -------------- A non-text attachment was scrubbed... Name: test-suite-aarch64.patch Type: text/x-patch Size: 957 bytes Desc: not available URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20140226/e777ffdc/attachment.bin>