[PATCH 0/4] PR c++/13403 and PR c++/15154: Fix function calling convention in gnuv3 ABI

[PATCH 0/4] PR c++/13403 and PR c++/15154: Fix function calling convention in gnuv3 ABI

[Siva Chandra]

This patch series addresses the two listed PRs. The gnuv3 ABI function
which decides whether a class value has to be passed by reference (or,
returned in a hidden first parameter) is currently incomplete. This
patch series completes it.

The way I have organised my patches is that:

- Patch 1/4 adds tests which illustrate what is missing. One test
fails, another passes.
- Patch 2/4 fixes the ABI to address the failing the test in 1/4.
- Patch 3/4 adds another failing test.
- Patch 4/4 fixes the ABI to address the failing test in 3/4.

If the patches go in in the same order as above, then the tree will
have one new failing test after 1/4 lands but before 2/4 lands. Is
that OK? If not, then I can merge 1 and 2 into one patch, and 3 and 4
into 1 patch. I have ordered it this way because I thought it might be
easier for the reviewers to see what works and what doesn't currently.

Re: [PATCH 0/4] PR c++/13403 and PR c++/15154: Fix function calling convention in gnuv3 ABI

[Siva Chandra]

On Thu, Sep 11, 2014 at 11:35 AM, Siva Chandra <sivachandra@google.com> wrote:
> This patch series addresses the two listed PRs. The gnuv3 ABI function
> which decides whether a class value has to be passed by reference (or,
> returned in a hidden first parameter) is currently incomplete. This
> patch series completes it.
>
> The way I have organised my patches is that:
>
> - Patch 1/4 adds tests which illustrate what is missing. One test
> fails, another passes.
> - Patch 2/4 fixes the ABI to address the failing the test in 1/4.
> - Patch 3/4 adds another failing test.
> - Patch 4/4 fixes the ABI to address the failing test in 3/4.
>
> If the patches go in in the same order as above, then the tree will
> have one new failing test after 1/4 lands but before 2/4 lands. Is
> that OK? If not, then I can merge 1 and 2 into one patch, and 3 and 4
> into 1 patch. I have ordered it this way because I thought it might be
> easier for the reviewers to see what works and what doesn't currently.

Ping.

For reference, links to patches 1/4 to 4/4 are as follows:

1/4 - https://sourceware.org/ml/gdb-patches/2014-09/msg00367.html
2/4 - https://sourceware.org/ml/gdb-patches/2014-09/msg00368.html
3/4 - https://sourceware.org/ml/gdb-patches/2014-09/msg00369.html
4/4 - https://sourceware.org/ml/gdb-patches/2014-09/msg00370.html

Re: [PATCH 0/4] PR c++/13403 and PR c++/15154: Fix function calling convention in gnuv3 ABI

[Doug Evans]

On Thu, Sep 11, 2014 at 11:35 AM, Siva Chandra <sivachandra@google.com> wrote:
> This patch series addresses the two listed PRs. The gnuv3 ABI function
> which decides whether a class value has to be passed by reference (or,
> returned in a hidden first parameter) is currently incomplete. This
> patch series completes it.
>
> The way I have organised my patches is that:
>
> - Patch 1/4 adds tests which illustrate what is missing. One test
> fails, another passes.
> - Patch 2/4 fixes the ABI to address the failing the test in 1/4.
> - Patch 3/4 adds another failing test.
> - Patch 4/4 fixes the ABI to address the failing test in 3/4.
>
> If the patches go in in the same order as above, then the tree will
> have one new failing test after 1/4 lands but before 2/4 lands. Is
> that OK? If not, then I can merge 1 and 2 into one patch, and 3 and 4
> into 1 patch. I have ordered it this way because I thought it might be
> easier for the reviewers to see what works and what doesn't currently.

That order is fine by me.

Re: [PATCH 0/4] PR c++/13403 and PR c++/15154: Fix function calling convention in gnuv3 ABI

[Siva Chandra]

On Sun, Sep 28, 2014 at 9:12 PM, Doug Evans <dje@google.com> wrote:
> On Thu, Sep 11, 2014 at 11:35 AM, Siva Chandra <sivachandra@google.com> wrote:
>> This patch series addresses the two listed PRs. The gnuv3 ABI function
>> which decides whether a class value has to be passed by reference (or,
>> returned in a hidden first parameter) is currently incomplete. This
>> patch series completes it.
>>
>> The way I have organised my patches is that:
>>
>> - Patch 1/4 adds tests which illustrate what is missing. One test
>> fails, another passes.
>> - Patch 2/4 fixes the ABI to address the failing the test in 1/4.
>> - Patch 3/4 adds another failing test.
>> - Patch 4/4 fixes the ABI to address the failing test in 3/4.
>>
>> If the patches go in in the same order as above, then the tree will
>> have one new failing test after 1/4 lands but before 2/4 lands. Is
>> that OK? If not, then I can merge 1 and 2 into one patch, and 3 and 4
>> into 1 patch. I have ordered it this way because I thought it might be
>> easier for the reviewers to see what works and what doesn't currently.
>
> That order is fine by me.

Pushed the patch set.

778811d5e7eb96b5ecb848033ffaa2df455a921e
82c48ac732edb0155288a93ef3dd39625ff2d2e1
2d1c107c1b8835f4e85c35320d8595a4a6fcaebe
ebb8ece2ef50ba3f86e8b4ab7a22a4c7734d114b