Thread (28 messages) 28 messages, 5 authors, 2022-08-06

Re: [GIT PULL] Block driver changes for 5.20-rc1

From: Jens Axboe <axboe@kernel.dk>
Date: 2022-08-02 22:24:21

On 8/2/22 3:50 PM, Linus Torvalds wrote:
On Tue, Aug 2, 2022 at 2:35 PM Jens Axboe [off-list ref] wrote:
quoted

As to testing, I'm going to punt that question to Hannes and Christoph,
as I have no way of testing that particular NVMe feature.
I can't test the *feature* either.

But dammit, I test two very different build configurations, and both
of them failed miserably on this file.

Don't you get it? That file DOES NOT EVEN COMPILE.

I refuse to have anything to do with a pull request that doesn't even
pass some very fundamental build requirements for me. That implies a
level of lack of testing that just makes me go "No way am I touching
that tree".
I can tell you that I always compile the whole damn thing, and this one
is no exception. The tree is also in for-next and has been for a long
time, both the drivers and drivers-post branch. The build bot has also
vetted both branches, individually, not just as the merged for-next.

I take it this is only happening on clang, which is why I haven't seen
it as I don't compile with clang. We can certainly add that to the usual
pre-flight/post-merge list, but I'm a bit surprised that clang isn't
being done by the build bots too.

If you want to make a clang build a hard requirement for any pull
request, then that should be explicit and not illicit outbursts if
that's just an implied assumption that it is being done. Really.

-- 
Jens Axboe
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help