check for invalid FP types in fsd #2216
Conversation
|
Thanks! Looks like the unit test (libutil/test/fsd.c) needs an update to pass a non-zero value for these tests that are expected to pass (Hmm, makes me wonder if @grondo had a use case for zero in mind?) and while in there, it would be good to add expected failures for the new bad values. We'll need a PR on https://github.com/flux-framework/rfc/blob/master/spec_23.adoc as well. |
|
Good point, I should have thought to add the test up front. I guess I'm ok with leaving zero ok if we have a use for that, maybe it's meaningful for something, or a key for "forever"? |
|
Maybe leave zero valid for now and we can fix it later if it's a loose end. |
Yeah, RFC14 defines a |
|
One builder failed in the kvs namespace test (looks like broker killed with -9). Appears unrelated, so I restarted. @trws if you could squash the incremental development this can go in IMHO. (fine to just squash down to one commit I think). |
|
Agreed, also realized I forgot to put in the check on the serialization side. Will do in the morning.
…Sent from my iPhone
On Jul 1, 2019, at 5:06 PM, Jim Garlick <notifications@github.com<mailto:notifications@github.com>> wrote:
One builder failed in the kvs namespace test (looks like broker killed with -9). Appears unrelated, so I restarted.
@trws<https://github.com/trws> if you could squash the incremental development this can go in IMHO.
(fine to just squash down to one commit I think).
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub<#2216?email_source=notifications&email_token=AAFBFNJJDZRXAOGTT3U3BWDP5KLWNA5CNFSM4H4TVRQ2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGODY7VS7Q#issuecomment-507468158>, or mute the thread<https://github.com/notifications/unsubscribe-auth/AAFBFNPOPSEIIRSTQED7BXTP5KLWNANCNFSM4H4TVRQQ>.
|
Adding a check to disallow NaNs and infinites in flux standard date format parsing and serialization in fsd. This came up while looking into flux-framework#2213. Also added a test for EINVAL on "NaNs" and "infinites", this also covers the various other forms of those, but I didn't see a reason to test strtod.
|
LGTM thanks! |
Adding a check to disallow zeros, NaNs and infinites in flux standard
duration format parsing in fsd. This came up while looking into #2213.
Just to see if we actually had issues with this, I tried a broker with shutdown grace duration of
NaNs, which hangs and has to be killed with signal 9 to die so, unlikely as this may be, may as well head it off.This probably needs an update in the RFC as well, but thought we could work it out here first.