[llvm] r243779 - IR: Add a broad bitcode compatibility test

[Sean Silva via llvm-commits]

Sorry for the delay, should be r244283 and r244284.

Thanks!

On Tue, Aug 4, 2015 at 3:51 PM, Sean Silva <chisophugis at gmail.com> wrote:

> Oh, it might be that the SVN is read-only for a moment: `[cfe-dev] Mailing
> list move NOW!`
>
> -- Sean Silva
>
> On Tue, Aug 4, 2015 at 3:50 PM, Sean Silva <chisophugis at gmail.com> wrote:
>
>> Could you try regenerating the patches? I seem to be having a hard time
>> applying them. For some reason there appear to be some CRLF in there maybe?
>>
>> -- Sean Silva
>>
>> On Tue, Aug 4, 2015 at 8:55 AM, Vedant Kumar <vsk at apple.com> wrote:
>>
>>> Would you mind committing these two patches for me?
>>>
>>> I don't have commit rights to llvm.
>>>
>>> thanks
>>> vedant
>>>
>>> > On Aug 3, 2015, at 7:46 PM, Sean Silva <chisophugis at gmail.com> wrote:
>>> >
>>> > LGTM. (I'm not an expert on the comdat feature but based on your
>>> process of following LangRef when writing the test I assume it covers the
>>> needed features).
>>> >
>>> > -- Sean Silva
>>> >
>>> > On Mon, Aug 3, 2015 at 9:19 AM, Vedant Kumar <vsk at apple.com> wrote:
>>> > I've responded to comments inline -- two patches attached.
>>> >
>>> > >>
>>> > >> -* The bitcode format produced by a X.Y release will be readable by
>>> all following
>>> > >> -  X.Z releases and the (X+1).0 release.
>>> > >> +* Additions and changes to the IR should be reflected in
>>> > >> +  ``test/Bitcode/compatibility.ll``.
>>> > >> +
>>> > >> +* The bitcode format produced by a X.Y release will be readable by
>>> all
>>> > >> +  following X.Z releases and the (X+1).0 release. To help ensure
>>> this, an X.Y
>>> > >> +  version of ``test/Bitcode/compatibility.ll`` should be assembled
>>> and
>>> > >> +  committed after each release.
>>> > >>
>>> > >> Please be a bit more specific here. I would like us to document
>>> explicitly the naming of the assembled files to avoid guesswork in the
>>> future.
>>> >
>>> > Here's the new language:
>>> >
>>> > > 508 * Additions and changes to the IR should be reflected in
>>> > > 509   ``test/Bitcode/compatibility.ll``.
>>> > > 510
>>> > > 511 * The bitcode format produced by a X.Y release will be readable
>>> by all
>>> > > 512   following X.Z releases and the (X+1).0 release.
>>> > > 513
>>> > > 514 * After each X.Y release, ``compatibility.ll`` must be copied to
>>> > > 515   ``compatibility-X.Y.ll``. The corresponding bitcode file
>>> should be assembled
>>> > > 516   using the X.Y build and committed as
>>> ``compatibility-X.Y.ll.bc``.
>>> >
>>> >
>>> >
>>> > >> + at comdat.samesize = global i32 0, comdat
>>> > >> +; CHECK: @comdat.samesize = global i32 0, comdat
>>> > >>
>>> > >> Is there a reason this is not covering usage of comdats with
>>> explicit comdat name?
>>> >
>>> > Not a good one. At one point I had mistakenly assumed that 3.6 lacked
>>> support for this.
>>> >
>>> > ``compatibility.ll`` has coverage for this, so I've copied it into the
>>> 3.6 test.
>>> >
>>> > [1] fix-combat-coverage.patch
>>> >
>>> > [Bitcode] Cover explicit comdat names in compatibility-3.6.ll
>>> >
>>> > [2] fix-devpolicy-language.patch
>>> >
>>> > [docs] Language about managing compatibility.ll made clearer
>>> >
>>> >
>>> >
>>>
>>>
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20150806/45b194da/attachment.html>