All of lore.kernel.org
 help / color / mirror / Atom feed
From: Randolph Sapp <rs@ti.com>
To: Denys Dmytriyenko <denis@denix.org>, Khem Raj <raj.khem@gmail.com>
Cc: <Openembedded-devel@lists.openembedded.org>, <afd@ti.com>,
	<detheridge@ti.com>, <reatmon@ti.com>
Subject: Re: [meta-oe][master][PATCH] vulkan-cts: add workaround for createMeshShaderMiscTestsEXT
Date: Tue, 18 Jun 2024 13:48:00 -0500	[thread overview]
Message-ID: <D23D8EROG3XR.95WKB5GSHMHP@ti.com> (raw)
In-Reply-To: <20240618182834.GA18231@denix.org>

On Tue Jun 18, 2024 at 1:28 PM CDT, Denys Dmytriyenko wrote:
> On Mon, Jun 17, 2024 at 06:41:57PM -0700, Khem Raj wrote:
> > On Mon, Jun 17, 2024 at 5:52 PM <rs@ti.com> wrote:
> > 
> > > From: Randolph Sapp <rs@ti.com>
> > >
> > > Remove the CXX flag "-O2" for GCC 13 and 14. There's a bug with ARM GCC
> > > that breaks the iteration of "types" in the createMeshShaderMiscTestsEXT
> > > function. This issue is not present for clang or x86_64 GCC 14.
>
> Is there an upstream bug submitted that tracks it?

Not that I know of, at least for GCC. Still trying to track down where that
should go and how to reproduce this in a way people will take it seriously.

About to have a PR in cts with two modifications to that function that also
solve the issue, though. Hoping to get more insight from them.

> > > It seems that the array is not initialized before the first iteration.
> > > In testing this can result in a random value being used. This can
> > > manifest in LINES type being processed twice, resulting in the following
> > > error:
> > >
> > > FATAL ERROR: Failed to initialize dEQP: Test case with non-unique name
> > > 'no_lines' added to group 'misc'.
> > >
> > > Signed-off-by: Randolph Sapp <rs@ti.com>
> > > ---
> > >  meta-oe/recipes-graphics/vk-gl-cts/vulkan-cts_1.3.7.3.bb | 3 +++
> > >  1 file changed, 3 insertions(+)
> > >
> > > diff --git a/meta-oe/recipes-graphics/vk-gl-cts/vulkan-cts_1.3.7.3.bb
> > > b/meta-oe/recipes-graphics/vk-gl-cts/vulkan-cts_1.3.7.3.bb
> > > index 1008a5fd4..7492fe9aa 100644
> > > --- a/meta-oe/recipes-graphics/vk-gl-cts/vulkan-cts_1.3.7.3.bb
> > > +++ b/meta-oe/recipes-graphics/vk-gl-cts/vulkan-cts_1.3.7.3.bb
> > > @@ -23,6 +23,9 @@ SRC_URI +=
> > > "file://0001-cmake-Define-WAYLAND_SCANNER-and-WAYLAND_PROTOCOLS_D.pat
> > >
> > >  TOOLCHAIN = "gcc"
> > >
> > > +# Workaround an optimization bug that breaks createMeshShaderMiscTestsEXT
> > > +OECMAKE_CXX_FLAGS:remove:toolchain-gcc = "-O2"
> > 
> > 
> > Let’s limit it to arm
> > And gcc using the needed overrides
> > 
> > >
> > > +
> > >  S = "${WORKDIR}/git"
> > >
> > >  REQUIRED_DISTRO_FEATURES = "vulkan"
> > > --
> > > 2.45.1
> > >
> > >



  reply	other threads:[~2024-06-18 18:48 UTC|newest]

Thread overview: 5+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2024-06-18  0:52 [meta-oe][master][PATCH] vulkan-cts: add workaround for createMeshShaderMiscTestsEXT rs
2024-06-18  1:41 ` Khem Raj
2024-06-18 18:28   ` Denys Dmytriyenko
2024-06-18 18:48     ` Randolph Sapp [this message]
2024-06-19 23:32       ` Randolph Sapp

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=D23D8EROG3XR.95WKB5GSHMHP@ti.com \
    --to=rs@ti.com \
    --cc=Openembedded-devel@lists.openembedded.org \
    --cc=afd@ti.com \
    --cc=denis@denix.org \
    --cc=detheridge@ti.com \
    --cc=raj.khem@gmail.com \
    --cc=reatmon@ti.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.