New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

next: Changes queue for v2018.06 #1361

Merged
merged 98 commits into from Jul 2, 2018

Conversation

Projects
None yet
5 participants
@lukego
Member

lukego commented Jun 19, 2018

This pulls the changes aggregated on max-next (#1350) and wingo-next (#1349) and should be good to go!

I reckon we can pull RaptorJIT onto next immediately after this release. It's working well for a bunch of people now (as discussed on Slack) but I don't think we need to squeeze it into this release. Seems worth a little extra time to e.g. review whether the README/documentation needs to be updated etc.

(Sorry for the relatively few cycles spent on upstreamery lately... It's a busy summer for me...)

eugeneia added some commits Jul 25, 2017

core.worker: include si_status in status
# Conflicts:
#	src/apps/inter/inter_test.lua
Merge fixes for initialization races from branch interlink into mcpring
# Conflicts:
#	src/apps/interlink/receiver.lua
#	src/apps/interlink/test.lua
#	src/apps/interlink/transmitter.lua
#	src/lib/interlink.lua
lib.ucd: supporting Unicode database for lib.xsd_regexp
General categories are only implemented for the ASCII range, the
predicates return false for all codepoints outside of ASCII.

For lib.xsd_regexp that means it can’t match category escapes on
non-ASCII input, i.e. it can’t fully handle Unicode input. It can
however apply all valid regular expressions to ASCII input.
Caveat: users of category escapes must restrict their input to ASCII.

In order to fix this, we would need (access to) a full Unicode database
in which we could look up codepoints and their attributes.

hepeng and others added some commits Apr 22, 2018

hepeng
fix a bug in tcp fields,
tcp.ack is the bit set in tcp flags
while tcp.ack_num should be the ack seq number.
Make "offset" arg to lseek a signed integer
Otherwise, passing a negative seek amount as a normal Lua number will
involve a cast from double to uint64.  In C it's undefined behavior when
a double outside the [0,2^64) range is cast to uint64.  In Lua we try to
additionally accomodate the range [-2^63, -1], but there is a bug on
x64-64 and might be a bug on other platforms:

  LuaJIT/LuaJIT#415

However it's cheaper to simply target an int64_t when you want to allow
negative numbers, as is our case, because you don't have to do assembly
heroics for it to do what you want.  The `tonumber` call in the return
of `linux.lua:lseek` indicates anyway that our range is not the full
64-bit range, so we might as well restrict instead to long rather than
ulong.
apps.interlink: avoid engine/appname kludge by attaching in :link()
By postponing attaching inter-process links to the :link() callback we can
avoid the kludge introduced to core.app that passes the app name to :new().
Semantics should be unchanged, although it might be a little less obvious when
the initialization happens.
@eugeneia

This comment has been minimized.

Member

eugeneia commented Jun 25, 2018

@lukego This is good to go except my lib.numa fixes (#1269) are broken, so we should revert
3120d7c.

@lukego

This comment has been minimized.

Member

lukego commented Jun 28, 2018

@eugeneia Revert pushed! Sorry about the wait.

@eugeneia

This comment has been minimized.

Member

eugeneia commented Jul 2, 2018

@lukego I am afraid I found another regression: I have bisected a SnabbNFV selftest failure down to d4c2eca (the merge commit, not the commit to be merged.) So d4c2eca should be reverted from the release until I can figure out what is wrong with it.

@eugeneia

This comment has been minimized.

Member

eugeneia commented Jul 2, 2018

Actually, never mind my last comment. The test was non-deterministic during bisect so the result is likely bogus. So this boiled down to the NFV selftest failing every other time on lugano-1. That is something to look into, but seems to not be affected by changes between v2018.04 and #1361. I am going forward with the release.

@eugeneia eugeneia merged commit cf4a5dc into master Jul 2, 2018

eugeneia added a commit that referenced this pull request Jul 2, 2018

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment