Update libbpf to v1.0.0 #216
Conversation
saschagrunert
force-pushed
the
libbpf-1.0.0
branch
12 times, most recently
from
0443742
to
e285a0b
Compare
|
This is great! Just as a reminder for folks migrating their application to libbpf v1, the strict mode API we added in #160 can be very useful to ensure a smooth transition |
|
Thanks @saschagrunert ! I'm really excited for this, hoping to have time to review and merge this by EOD tomorrow and a new release sometime next week once we're sure we've removed everything that's deprecated and firmly standardized error checking the way 1.0 does it. I'll also share publicly our roadmap towards libbpfgo 1.0 which will certainly have to include a full implementation of the libbpf API. I gotta setup a public office hours so we (all contributors/maintainers) can discuss that and other exciting things we can do such as being part of the libbpf project more officially. Will keep you posted! |
And things that don't need documentation can also be discussed in Slack (#libbpfgo), as long as we also keep track of things in github (Always). |
As Parca Agent maintainers, we would be more than happy to give a hand on this effort.
❤️ It will be amazing to have office hours! |
|
I'm presently going through libbpfgo to make sure we're not using any now removed APIs and are handling errors correctly. There's also a bunch of new APIs and things that we can do differently in 1.0 that'll i'm going to be creating issues for but won't hold up this PR (but will be before the next libbpfgo release). |
❤️
Will let you know very soon! |
|
Alright I've opened #225 which we'll merge at the same time as this one. For some reason github doesn't let me tag people out of the aqua org for review but @kakkoyun @javierhonduco @saschagrunert your feedback is always greatly appreciated! |
| @@ -1031,7 +1031,7 @@ func (it *BPFMapIterator) Next() bool { | |||
| nextPtr := unsafe.Pointer(&next[0]) | |||
|
|
|||
| errC, err := C.bpf_map_get_next_key(it.b.fd, prevPtr, nextPtr) | |||
| if errno, ok := err.(syscall.Errno); errC == -1 && ok && errno == C.ENOENT { | |||
| if errno, ok := err.(syscall.Errno); errC == -2 && ok && errno == C.ENOENT { | |||
There was a problem hiding this comment.
That's the return code for the function call, the iterator test failed because of this change.
There was a problem hiding this comment.
Got it! I see now. i guess it's redundant to check errno == C.ENOENT and errC = -2. Do you think we should simplify to just the following?:
| if errno, ok := err.(syscall.Errno); errC == -2 && ok && errno == C.ENOENT { | |
| if errC == -2 { |
There was a problem hiding this comment.
This is actually outside the scope of this PR, so I'd say leave it for now and we can reassess before the next release
Updates libbpf to the latest release: https://github.com/libbpf/libbpf/releases/tag/v1.0.0 Signed-off-by: Sascha Grunert <sgrunert@redhat.com>
saschagrunert
force-pushed
the
libbpf-1.0.0
branch
from
e285a0b
to
aaaa25c
Compare
|
Thanks again @saschagrunert, Ill be looking to cut a release by middle of next week. |
Updates libbpf to the latest release: https://github.com/libbpf/libbpf/releases/tag/v1.0.0