Re: [dpdk-ci] [dpdk-dev] [pull-request] dpdk-next-net-eventdev - 21.05 - PRE-RC1

Re: [dpdk-ci] [dpdk-dev] [pull-request] dpdk-next-net-eventdev - 21.05 - PRE-RC1

[David Marchand]

On Tue, Apr 13, 2021 at 12:12 AM Thomas Monjalon <thomas@monjalon.net> wrote:
>
> 12/04/2021 15:20, Jerin Jacob Kollanukkaran:
> >   http://dpdk.org/git/next/dpdk-next-eventdev
>
> Pulled, thanks.
>
> Note few changes in titles like uppercases for acronyms,
> or "add support" simplified as "support",
> and release notes moved in the right place.

The ABI check now reports an error on event/dlb.
The reason is that the event/dlb driver has been removed, and so the
check complains about a missing dump.
This will have to be fixed quickly or reverted.


This has been missed by multiple people, so trying to understand why.

The ABI check should have caught this when run by maintainers (/me
looks at Thomas and Jerin).

The CI should have caught it too.
But, v1 did not apply.
For v2, I can see a doc generation issue reported by Intel CI that I
can't reproduce, so it could be just noise.
I can't find reports for Travis or GHA and I could not find in the
robot logs why the series_15708 branch was not created.

Looking at UNH reports:
http://mails.dpdk.org/archives/test-report/2021-March/182956.html
But looking at the log:
[2713/2716] Compiling C object
'drivers/a715181@@rte_event_octeontx2@sta/meson-generated_.._rte_event_octeontx2.pmd.c.o'.
[2714/2716] Linking static target drivers/librte_event_octeontx2.a.
[2715/2716] Generating rte_event_octeontx2.sym_chk with a meson_exe.py
custom command.
[2716/2716] Linking target drivers/librte_event_octeontx2.so.21.1.
Error: cannot find librte_event_dlb.dump in
/home-local/jenkins-local/jenkins-agent/workspace/Ubuntu18.04-Compile-DPDK-ABI/dpdk/build-gcc-shared/install

Is this something that has been fixed since then?

I don't have the main branch/recent series status from UNH, but at
least GHA and Travis are now complaining about ABI.


-- 
David Marchand

Re: [dpdk-ci] [dpdk-dev] [pull-request] dpdk-next-net-eventdev - 21.05 - PRE-RC1

[Thomas Monjalon]

13/04/2021 09:15, David Marchand:
> On Tue, Apr 13, 2021 at 12:12 AM Thomas Monjalon <thomas@monjalon.net> wrote:
> >
> > 12/04/2021 15:20, Jerin Jacob Kollanukkaran:
> > >   http://dpdk.org/git/next/dpdk-next-eventdev
> >
> > Pulled, thanks.
> >
> > Note few changes in titles like uppercases for acronyms,
> > or "add support" simplified as "support",
> > and release notes moved in the right place.
> 
> The ABI check now reports an error on event/dlb.
> The reason is that the event/dlb driver has been removed, and so the
> check complains about a missing dump.
> This will have to be fixed quickly or reverted.
> 
> 
> This has been missed by multiple people, so trying to understand why.
> 
> The ABI check should have caught this when run by maintainers (/me
> looks at Thomas and Jerin).
> 
> The CI should have caught it too.
> But, v1 did not apply.
> For v2, I can see a doc generation issue reported by Intel CI that I
> can't reproduce, so it could be just noise.
> I can't find reports for Travis or GHA and I could not find in the
> robot logs why the series_15708 branch was not created.
> 
> Looking at UNH reports:
> http://mails.dpdk.org/archives/test-report/2021-March/182956.html
> But looking at the log:
> [2713/2716] Compiling C object
> 'drivers/a715181@@rte_event_octeontx2@sta/meson-generated_.._rte_event_octeontx2.pmd.c.o'.
> [2714/2716] Linking static target drivers/librte_event_octeontx2.a.
> [2715/2716] Generating rte_event_octeontx2.sym_chk with a meson_exe.py
> custom command.
> [2716/2716] Linking target drivers/librte_event_octeontx2.so.21.1.
> Error: cannot find librte_event_dlb.dump in
> /home-local/jenkins-local/jenkins-agent/workspace/Ubuntu18.04-Compile-DPDK-ABI/dpdk/build-gcc-shared/install
> 
> Is this something that has been fixed since then?
> 
> I don't have the main branch/recent series status from UNH, but at
> least GHA and Travis are now complaining about ABI.

Sorry about that, this is because I stupidly thought I could fix it
locally without thinking about the CI.
I am going to send a patch for devtools/libabigail.abignore.

Re: [dpdk-ci] [dpdk-dev] [pull-request] dpdk-next-net-eventdev - 21.05 - PRE-RC1

[Jerin Jacob]

On Tue, Apr 13, 2021 at 12:46 PM David Marchand
<david.marchand@redhat.com> wrote:
>
> On Tue, Apr 13, 2021 at 12:12 AM Thomas Monjalon <thomas@monjalon.net> wrote:
> >
> > 12/04/2021 15:20, Jerin Jacob Kollanukkaran:
> > >   http://dpdk.org/git/next/dpdk-next-eventdev
> >
> > Pulled, thanks.
> >
> > Note few changes in titles like uppercases for acronyms,
> > or "add support" simplified as "support",
> > and release notes moved in the right place.
>
> The ABI check now reports an error on event/dlb.
> The reason is that the event/dlb driver has been removed, and so the
> check complains about a missing dump.
> This will have to be fixed quickly or reverted.
>
>
> This has been missed by multiple people, so trying to understand why.


I was running the following script[1] to detect ABI issues.
Since the "./devtools/test-meson-builds.sh" did not return non zero value or
the error print was "Error: cannot find librte_event_dlb.dump", It is
missed from my side.

@David Marchand  @Thomas Monjalon Could you share the snippet you are
using for detecting the ABI issue.


------------------------
# ABI check
DPDK_ABI_REF_VERSION=v20.11 DPDK_ABI_REF_DIR=/tmp bash
./devtools/test-meson-builds.sh 1> /tmp/build.log 2> /tmp/build.log
if [ $? -ne 0 ]; then
        echo "ABI check failed"
        exit
fi

grep "Error: ABI issue reported" /tmp/build.log
if [ $? -eq 0 ]; then
        echo "ABI issue"
        exit
fi

-------------------------------------------------------
>
> The ABI check should have caught this when run by maintainers (/me
> looks at Thomas and Jerin).

Sorry for that :-(

Re: [dpdk-ci] [dpdk-dev] [pull-request] dpdk-next-net-eventdev - 21.05 - PRE-RC1

[Thomas Monjalon]

13/04/2021 10:54, Jerin Jacob:
> On Tue, Apr 13, 2021 at 12:46 PM David Marchand
> <david.marchand@redhat.com> wrote:
> >
> > On Tue, Apr 13, 2021 at 12:12 AM Thomas Monjalon <thomas@monjalon.net> wrote:
> > >
> > > 12/04/2021 15:20, Jerin Jacob Kollanukkaran:
> > > >   http://dpdk.org/git/next/dpdk-next-eventdev
> > >
> > > Pulled, thanks.
> > >
> > > Note few changes in titles like uppercases for acronyms,
> > > or "add support" simplified as "support",
> > > and release notes moved in the right place.
> >
> > The ABI check now reports an error on event/dlb.
> > The reason is that the event/dlb driver has been removed, and so the
> > check complains about a missing dump.
> > This will have to be fixed quickly or reverted.
> >
> >
> > This has been missed by multiple people, so trying to understand why.
> 
> 
> I was running the following script[1] to detect ABI issues.
> Since the "./devtools/test-meson-builds.sh" did not return non zero value or
> the error print was "Error: cannot find librte_event_dlb.dump", It is
> missed from my side.
> 
> @David Marchand  @Thomas Monjalon Could you share the snippet you are
> using for detecting the ABI issue.

I do like you: simply run test-meson-builds.sh
And yes I saw the error, and I don't know why I thought it was OK!
We are humans :)


> ------------------------
> # ABI check
> DPDK_ABI_REF_VERSION=v20.11 DPDK_ABI_REF_DIR=/tmp bash
> ./devtools/test-meson-builds.sh 1> /tmp/build.log 2> /tmp/build.log
> if [ $? -ne 0 ]; then
>         echo "ABI check failed"
>         exit
> fi
> 
> grep "Error: ABI issue reported" /tmp/build.log
> if [ $? -eq 0 ]; then
>         echo "ABI issue"
>         exit
> fi
> 
> -------------------------------------------------------
> >
> > The ABI check should have caught this when run by maintainers (/me
> > looks at Thomas and Jerin).
> 
> Sorry for that :-(

Re: [dpdk-ci] [dpdk-dev] [pull-request] dpdk-next-net-eventdev - 21.05 - PRE-RC1

[David Marchand]

On Tue, Apr 13, 2021 at 10:55 AM Jerin Jacob <jerinjacobk@gmail.com> wrote:
> I was running the following script[1] to detect ABI issues.
> Since the "./devtools/test-meson-builds.sh" did not return non zero value or
> the error print was "Error: cannot find librte_event_dlb.dump", It is
> missed from my side.
>
> @David Marchand  @Thomas Monjalon Could you share the snippet you are
> using for detecting the ABI issue.
>
>
> ------------------------
> # ABI check
> DPDK_ABI_REF_VERSION=v20.11 DPDK_ABI_REF_DIR=/tmp bash

Ah ok, this is because event/dlb did not exist in 20.11.
Running against 21.02, you should get the error.


-- 
David Marchand

Re: [dpdk-ci] [dpdk-dev] [pull-request] dpdk-next-net-eventdev - 21.05 - PRE-RC1

[Thomas Monjalon]

13/04/2021 11:07, David Marchand:
> On Tue, Apr 13, 2021 at 10:55 AM Jerin Jacob <jerinjacobk@gmail.com> wrote:
> > I was running the following script[1] to detect ABI issues.
> > Since the "./devtools/test-meson-builds.sh" did not return non zero value or
> > the error print was "Error: cannot find librte_event_dlb.dump", It is
> > missed from my side.
> >
> > @David Marchand  @Thomas Monjalon Could you share the snippet you are
> > using for detecting the ABI issue.
> >
> >
> > ------------------------
> > # ABI check
> > DPDK_ABI_REF_VERSION=v20.11 DPDK_ABI_REF_DIR=/tmp bash
> 
> Ah ok, this is because event/dlb did not exist in 20.11.

No it was added in 20.11.

> Running against 21.02, you should get the error.

Yes we should always run the ABI check against the latest release.
It means you must upgrade DPDK_ABI_REF_VERSION after each release.

Re: [dpdk-ci] [dpdk-dev] [pull-request] dpdk-next-net-eventdev - 21.05 - PRE-RC1

[David Marchand]

On Tue, Apr 13, 2021 at 11:07 AM David Marchand
<david.marchand@redhat.com> wrote:
>
> On Tue, Apr 13, 2021 at 10:55 AM Jerin Jacob <jerinjacobk@gmail.com> wrote:
> > I was running the following script[1] to detect ABI issues.
> > Since the "./devtools/test-meson-builds.sh" did not return non zero value or
> > the error print was "Error: cannot find librte_event_dlb.dump", It is
> > missed from my side.
> >
> > @David Marchand  @Thomas Monjalon Could you share the snippet you are
> > using for detecting the ABI issue.
> >
> >
> > ------------------------
> > # ABI check
> > DPDK_ABI_REF_VERSION=v20.11 DPDK_ABI_REF_DIR=/tmp bash
>
> Ah ok, this is because event/dlb did not exist in 20.11.
> Running against 21.02, you should get the error.

Clicked send too quickly...

No, it did exist.

$ DPDK_ABI_REF_VERSION=v20.11 ./devtools/test-meson-builds.sh
ninja: Entering directory `/home/dmarchan/builds/build-gcc-static'
ninja: no work to do.
ninja: Entering directory `/home/dmarchan/builds/build-gcc-shared'
ninja: no work to do.
Error: cannot find librte_event_dlb.dump in
/home/dmarchan/builds/build-gcc-shared/install

Hum... the next reason I see would be that your reference does not
have event/dlb which is surprising with default configuration.
$ ls $DPDK_ABI_REF_DIR/*/*/dump/*dlb.dump
/home/dmarchan/abi/v20.11/build-clang-shared/dump/librte_event_dlb.dump
 /home/dmarchan/abi/v21.02/build-clang-shared/dump/librte_event_dlb.dump
/home/dmarchan/abi/v20.11/build-gcc-shared/dump/librte_event_dlb.dump
  /home/dmarchan/abi/v21.02/build-gcc-shared/dump/librte_event_dlb.dump


-- 
David Marchand

Re: [dpdk-ci] [dpdk-dev] [pull-request] dpdk-next-net-eventdev - 21.05 - PRE-RC1

[Aaron Conole]

David Marchand <david.marchand@redhat.com> writes:

> On Tue, Apr 13, 2021 at 12:12 AM Thomas Monjalon <thomas@monjalon.net> wrote:
>>
>> 12/04/2021 15:20, Jerin Jacob Kollanukkaran:
>> >   http://dpdk.org/git/next/dpdk-next-eventdev
>>
>> Pulled, thanks.
>>
>> Note few changes in titles like uppercases for acronyms,
>> or "add support" simplified as "support",
>> and release notes moved in the right place.
>
> The ABI check now reports an error on event/dlb.
> The reason is that the event/dlb driver has been removed, and so the
> check complains about a missing dump.
> This will have to be fixed quickly or reverted.
>
>
> This has been missed by multiple people, so trying to understand why.
>
> The ABI check should have caught this when run by maintainers (/me
> looks at Thomas and Jerin).
>
> The CI should have caught it too.
> But, v1 did not apply.
> For v2, I can see a doc generation issue reported by Intel CI that I
> can't reproduce, so it could be just noise.
> I can't find reports for Travis or GHA and I could not find in the
> robot logs why the series_15708 branch was not created.

I need to go back in the logs - there were a few cases during a
change-over that the robot wasn't running right.

> Looking at UNH reports:
> http://mails.dpdk.org/archives/test-report/2021-March/182956.html
> But looking at the log:
> [2713/2716] Compiling C object
> 'drivers/a715181@@rte_event_octeontx2@sta/meson-generated_.._rte_event_octeontx2.pmd.c.o'.
> [2714/2716] Linking static target drivers/librte_event_octeontx2.a.
> [2715/2716] Generating rte_event_octeontx2.sym_chk with a meson_exe.py
> custom command.
> [2716/2716] Linking target drivers/librte_event_octeontx2.so.21.1.
> Error: cannot find librte_event_dlb.dump in
> /home-local/jenkins-local/jenkins-agent/workspace/Ubuntu18.04-Compile-DPDK-ABI/dpdk/build-gcc-shared/install
>
> Is this something that has been fixed since then?
>
> I don't have the main branch/recent series status from UNH, but at
> least GHA and Travis are now complaining about ABI.