[llvm] r230803 - Revert test case until it can be fixed

David Blaikie dblaikie at gmail.com
Sun Mar 1 08:01:01 PST 2015


On Mar 1, 2015 6:52 AM, "Bill Schmidt" <wschmidt at linux.vnet.ibm.com> wrote:
>
> On Fri, 2015-02-27 at 14:50 -0800, David Blaikie wrote:
> >
> >
> > On Fri, Feb 27, 2015 at 2:31 PM, Bill Schmidt
> > <wschmidt at linux.vnet.ibm.com> wrote:
> >         Author: wschmidt
> >         Date: Fri Feb 27 16:31:14 2015
> >         New Revision: 230803
> >
> >         URL: http://llvm.org/viewvc/llvm-project?rev=230803&view=rev
> >         Log:
> >         Revert test case until it can be fixed
> >
> > Usually if  a test is going to be reverted, so should the change (just
> > revert the whole commit, fix it, recommit) - rather than leaving the
> > functionality untested in-tree.
>
> Yes, I'm aware of that, but in this case it was clear that I had just
> effed the test case due to your changes and wanted to quickly shut up
> the build bots

Reverting still does a pretty good job of during up the buildbots and is
about as timely to execute. Helps maintain the 'code is tested' line and
avoid reviewers having to check that the test for resubmitted at some point
(did it? A follow up email to this review mentioning the revision where
this was resubmitted would be helpful)

> till I could get through a 10-minute fix/test cycle.

What it is it in your cycle that takes ten minutes? If you are only
modifying one test file then you can be fairly sure it won't affect other
tests, so you can just run that test in isolation. Lit can take the project
relative test path (or parent directory path) to run individual tests. Can
be quite handy to iterate faster rather than waiting for complete test
cycles.

> If
> it had been a more pervasive issue I would of course have reverted the
> fix as well.
>
> Bill
> >
> > In this case it looks like you just need to update for my recent
> > explicit type change for load and gep.
> >
> >
> >         Removed:
> >             llvm/trunk/test/CodeGen/PowerPC/pr22711.ll
> >
> >         Removed: llvm/trunk/test/CodeGen/PowerPC/pr22711.ll
> >         URL:
> >
http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/PowerPC/pr22711.ll?rev=230802&view=auto
> >
 ==============================================================================
> >         --- llvm/trunk/test/CodeGen/PowerPC/pr22711.ll (original)
> >         +++ llvm/trunk/test/CodeGen/PowerPC/pr22711.ll (removed)
> >         @@ -1,65 +0,0 @@
> >         -; Verify that the .toc section is aligned on an 8-byte
> >         boundary.
> >         -
> >         -; RUN: llc < %s -mtriple=powerpc64-unknown-linux-gnu
> >         -mcpu=pwr8 -filetype=obj -o - | llvm-readobj --sections |
> >         FileCheck %s
> >         -
> >         -define void @test(i32* %a) {
> >         -entry:
> >         -  %a.addr = alloca i32*, align 8
> >         -  store i32* %a, i32** %a.addr, align 8
> >         -  %0 = load i32** %a.addr, align 8
> >
> > %0 = load i32*, i32** %a.addr, align 8
> >
> > (& similar changes for all the other loads)
> >
> >         -  %incdec.ptr = getelementptr inbounds i32* %0, i32 1
> >
> > %incdec.ptr = getelementptr inbounds i32, i32* %0, i32 1
> >
> >         -  store i32* %incdec.ptr, i32** %a.addr, align 8
> >         -  %1 = load i32* %0, align 4
> >         -  switch i32 %1, label %sw.epilog [
> >         -    i32 17, label %sw.bb
> >         -    i32 13, label %sw.bb1
> >         -    i32 11, label %sw.bb2
> >         -    i32 7, label %sw.bb3
> >         -    i32 5, label %sw.bb4
> >         -    i32 3, label %sw.bb5
> >         -    i32 2, label %sw.bb6
> >         -  ]
> >         -
> >         -sw.bb:                                            ; preds = %
> >         entry
> >         -  %2 = load i32** %a.addr, align 8
> >         -  store i32 2, i32* %2, align 4
> >         -  br label %sw.epilog
> >         -
> >         -sw.bb1:                                           ; preds = %
> >         entry
> >         -  %3 = load i32** %a.addr, align 8
> >         -  store i32 3, i32* %3, align 4
> >         -  br label %sw.epilog
> >         -
> >         -sw.bb2:                                           ; preds = %
> >         entry
> >         -  %4 = load i32** %a.addr, align 8
> >         -  store i32 5, i32* %4, align 4
> >         -  br label %sw.epilog
> >         -
> >         -sw.bb3:                                           ; preds = %
> >         entry
> >         -  %5 = load i32** %a.addr, align 8
> >         -  store i32 7, i32* %5, align 4
> >         -  br label %sw.epilog
> >         -
> >         -sw.bb4:                                           ; preds = %
> >         entry
> >         -  %6 = load i32** %a.addr, align 8
> >         -  store i32 11, i32* %6, align 4
> >         -  br label %sw.epilog
> >         -
> >         -sw.bb5:                                           ; preds = %
> >         entry
> >         -  %7 = load i32** %a.addr, align 8
> >         -  store i32 13, i32* %7, align 4
> >         -  br label %sw.epilog
> >         -
> >         -sw.bb6:                                           ; preds = %
> >         entry
> >         -  %8 = load i32** %a.addr, align 8
> >         -  store i32 17, i32* %8, align 4
> >         -  br label %sw.epilog
> >         -
> >         -sw.epilog:                                        ; preds = %
> >         entry, %sw.bb6, %sw.bb5, %sw.bb4, %sw.bb3, %sw.bb2, %sw.bb1, %
> >         sw.bb
> >         -  ret void
> >         -}
> >         -
> >         -; CHECK: Name: .toc
> >         -; CHECK: AddressAlignment: 8
> >         -; CHECK: Name: .rela.toc
> >         -
> >
> >
> >         _______________________________________________
> >         llvm-commits mailing list
> >         llvm-commits at cs.uiuc.edu
> >         http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
> >
> >
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20150301/0a3e2632/attachment.html>


More information about the llvm-commits mailing list