workload/ycsb: package failed under stress #34004
Comments
cockroach-teamcity
added
C-test-failure
|
Using this as an anchor for all the other issues like it. I can repro this locally using I think |
|
That's probably it, GOFLAGS is new: If this is true, we should "just" require Go 1.11 and stop passing GOFLAGS explicitly. Or we must prefix every invocation of @knz what do you think is the sane option here? |
|
I thought peter had a PR to bump the req to 1.11? I'd vote for that then we can remove the explicit GOFLAGS. |
|
We do require go1.11. I wasn’t aware of this new GOFLAGS behavior or how it
would interact with nightly race tests.
…On Tue, Jan 15, 2019 at 5:34 AM kena ***@***.***> wrote:
I thought peter had a PR to bump the req to 1.11? I'd vote for that then
we can remove the explicit GOFLAGS.
—
You are receiving this because you are subscribed to this thread.
Reply to this email directly, view it on GitHub
<#34004 (comment)>,
or mute the thread
<https://github.com/notifications/unsubscribe-auth/AF6f99vC9nLNNWYhkIB1xn-PSh4ydtKlks5vDa6kgaJpZM4aAad7>
.
|
|
Something about this is still janky. I just tried this out (i.e. not passing Perhaps someone with more Makefile knowledge can look at this. @petermattis? |
|
(My preference right now would be making sure that the Makefile overrides GOFLAGS with an empty value initially, i.e. we don't let anything from the env slip in). BTW I thought the non-working case that I describe above should work, according to https://ftp.gnu.org/old-gnu/Manuals/make-3.79.1/html_chapter/make_6.html#SEC68, though I don't even know whether I understood that correctly or whether it applies to our version of |
|
I’ll take a look today. Can you send me a branch with what you tried so I
can make new and different mistakes?
…On Tue, Jan 15, 2019 at 6:06 AM Tobias Grieger ***@***.***> wrote:
(My preference right now would be making sure that the Makefile overrides
GOFLAGS with an empty value initially, i.e. we don't let anything from the
env slip in).
BTW I thought the non-working case that I describe above should work,
according to
https://ftp.gnu.org/old-gnu/Manuals/make-3.79.1/html_chapter/make_6.html#SEC68,
though I don't even know whether I understood that correctly or whether it
applies to our version of make.
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
<#34004 (comment)>,
or mute the thread
<https://github.com/notifications/unsubscribe-auth/AF6f9yNIdPqgGK1oZGNY6KtSv_8PI9HSks5vDbYhgaJpZM4aAad7>
.
|
|
Thanks @petermattis. I didn't save the branch, unfortunately but there really wasn't much to it. |
|
Note that |
|
I looked at that but it seemed nasty, also the general problem persists -- if you have GOFLAGS in your env, things will be weird and that's not good. It really seems like we want to make sure you don't pass GOFLAGS in through the environment. (Or rather, that |
|
GNU Make 3.82 provides a handy undefine function. But the make version on my computer is 3.81. Grrr. We invoke |
As of go1.11, `go` interprets the `GOFLAGS` env variable and errors out if certain flags are duplicated in both the environment variable and on the command line (e.g. `-race`). Fixes cockroachdb#34004 Release note: None
34013: build: avoid duplication of GOFLAGS for nightly stress r=jordanlewis,tbg a=petermattis Clear the `GOFLAGS` env variable for the nightly stress invocations. As of go1.11, `go` interprets that variable and errors out if certain flags are duplicated in both the environment variable and on the command line (e.g. `-race`). Fixes #34004 Release note: None Co-authored-by: Peter Mattis <petermattis@gmail.com>
SHA: https://github.com/cockroachdb/cockroach/commits/0c7e491a76142a67750649facbce798442f74d25
Parameters:
To repro, try:
Failed test: https://teamcity.cockroachdb.com/viewLog.html?buildId=1094320&tab=buildLog
The text was updated successfully, but these errors were encountered: