Bug 212638

Summary: emulators/virtualbox-ose-kmod & x11/nvidia-driver no longer build with NO_MODULES
Product: Ports & Packages Reporter: Alastair Hogge <agh>
Component: Individual Port(s)Assignee: Virtualbox Team (Nobody) <vbox>
Status: Open ---    
Severity: Affects Some People CC: agh, bdrewery, danfe, grahamperrin, kevans, koobs, vbox, w.schwarzenfeld
Priority: --- Keywords: needs-qa
Version: LatestFlags: koobs: maintainer-feedback? (kevans)
koobs: maintainer-feedback? (vbox)
koobs: maintainer-feedback? (danfe)
Hardware: amd64   
OS: Any   
Attachments:
Description Flags
virtualbox-ose-kmod-5.1.6 poudriere build log
none
nvidia-driver-390.87_3 build log
none
virtualbox-ose-kmod-5.2.32_1 build log none

Description Alastair Hogge 2016-09-13 02:53:00 UTC
Created attachment 174702 [details]
virtualbox-ose-kmod-5.1.6 poudriere build log

Hi,

Ports that require building kernel modules are no longer successfully building.  ~3 months ago pkg builds of emulators/virtualbox-ose-kmod & x11/nvidia-driver would build successfully on CURRENT with NO_MODULES=YES in make.conf.  With ports at r421970 & base at r305515 the build fails with "make[2]: make[2]: don't know how to make device_if.h. Stop"

Is this now intentional?
Comment 1 Walter Schwarzenfeld freebsd_triage 2018-01-14 03:21:51 UTC
We have 5.2.4. Seems overcome by events.
Comment 2 Kubilay Kocak freebsd_committer freebsd_triage 2019-06-27 13:07:43 UTC
Version being newer is insufficient evidence to close OBE. Positive evidence that it is not or lo longer reproducible is necessary, or feedback timeout (after a request).

We have a report via IRC that a test with NO_MODULES no longer fails, waiting on explicit comment to confirm that.
Comment 3 Alastair Hogge 2019-08-22 06:25:04 UTC
Created attachment 206780 [details]
nvidia-driver-390.87_3 build log

Still failing on 13-CURRENT r351006 and ports at r509577
Comment 4 Alastair Hogge 2019-08-22 06:25:52 UTC
Created attachment 206781 [details]
virtualbox-ose-kmod-5.2.32_1 build log

Still failing on 13-CURRENT r351006 and ports at r509577
Comment 5 Kubilay Kocak freebsd_committer freebsd_triage 2020-05-18 01:58:43 UTC
(In reply to Alastair Hogge from comment #4)

Is this still an issue Alastair?

@Kyle Are you able to easily test reproduction of this?

^Triage: 

Close this as Unable to Reproduce, requesting re-open & steps to reproduce from reporters/previous confirmers if still reproducible there.

Otherwise close Feedback Timeout in 2 weeks, requesting re-open if still reproducible
Comment 6 Kubilay Kocak freebsd_committer freebsd_triage 2020-05-18 01:59:20 UTC
^Triage: Request feedback from nvidia-driver maintainer
Comment 7 Kyle Evans freebsd_committer 2020-05-18 02:19:19 UTC
Given the timeline, I think r303195 by bdrewery@ is the likely cause:

```
Don't run find(1) for __MPATH with NO_MODULES set.

It's a waste of time when it won't be used.

Submitted by:	bde
MFC after:	3 days
```

That said, it really looks like the fact that port kmod builds being able to succeed prior to that was technically a bug. I'm CC'ing bdrewery@, as he'd be better to consult on whether these being broken should be expected or not.

If they shouldn't work, then perhaps we can improve USES=kmod to try and detect NO_MODULES being defined and set IGNORE, though that's a little more complicated since it could perhaps be set in src.conf or src-env.conf instead.