<JRubyGithub>
[jruby] mkristian pushed 1 new commit to test-merge-1_7: http://git.io/vIgpL
<JRubyGithub>
jruby/test-merge-1_7 833d8e8 Christian Meier: fix merge issue on test_file.rb...
JRubyGithub has left #jruby [#jruby]
deobalds has joined #jruby
lance|afk is now known as lanceball
colinsurprenant has joined #jruby
iamjarvo has joined #jruby
tcrawley is now known as tcrawley-away
tcrawley-away is now known as tcrawley
tenderlove has quit [Remote host closed the connection]
tenderlove has joined #jruby
enebo has joined #jruby
rsim has joined #jruby
oblutak has joined #jruby
rsim has quit [Client Quit]
JRubyGithub has joined #jruby
<JRubyGithub>
[jruby] mkristian force-pushed test-dir-globs-on-uri-classloader from ba6d67c to 3b97172: http://git.io/vkVCI
<JRubyGithub>
jruby/test-dir-globs-on-uri-classloader ba68e39 Christian Meier: find directory entries inside the jar if there is no .jrubydir file...
<JRubyGithub>
jruby/test-dir-globs-on-uri-classloader 481af63 Christian Meier: added current directory to load_path for spec:profiler
<JRubyGithub>
jruby/test-dir-globs-on-uri-classloader 82fbf83 Christian Meier: some integration tests do see uri:classloader: for jruby-rack applications
JRubyGithub has left #jruby [#jruby]
JRubyGithub has joined #jruby
<JRubyGithub>
[jruby] enebo pushed 1 new commit to master: http://git.io/vI2sP
<JRubyGithub>
jruby/master 1baad69 Thomas E. Enebo: Update for released deps
JRubyGithub has left #jruby [#jruby]
<bbrowning>
enebo: are there any tests that verify "jruby -0" behavior other than the failing one in test_command_line_switches.rb?
<bbrowning>
seems like that's broken on at least some level in 9k
<bbrowning>
of course, "broken" here is relative to 1.7 - looks like 1.7 was also broken in comparison to MRI
<bbrowning>
but the test passed, at least
colinsurprenant has quit [Quit: colinsurprenant]
<enebo>
bbrowning: I broke this months ago
<enebo>
bbrowning: I will attempt a different fix but look at: Ruby.addGetsLoop
<bbrowning>
is that the snippet you're referring to that mri22 gets wrong?
<enebo>
bbrowning: oh yeah
<bbrowning>
jruby has some other odd behavior in that setting the record separator "-0" on the command line sets both the input and output record separators
<bbrowning>
but with mri -0 only controls the input separator
<bbrowning>
so "-054" causes jruby to separate puts statements with "," instead of "\n"
<bbrowning>
but 1.7 had that same issue
erikhatcher has quit [Quit: erikhatcher]
<bbrowning>
anyway, I was only looking at this because that one test in slow_suites is failing
<enebo>
bbrowning: yeah there is the printing logic but I did not think that happened from -)
<enebo>
err -)
<enebo>
haha -0
<bbrowning>
-0 should not control the printing logic, but it does
<enebo>
err processLineEndings
<enebo>
but processLineEndings is not affected by -)
<bbrowning>
enebo: do any CI jobs run the rubyspecs?
<bbrowning>
hah
<enebo>
bbrowning: I don’t think so but I thought we had the longer slowers ones on cloudbees
<bbrowning>
bees doesn't run anything that travis doesn't afaik
<enebo>
bbrowning: then I don’t think so
<bbrowning>
ahh I see there are jobs for some slower ones but they were disabled a while back for some reason
<enebo>
bbrowning: I thought the main problem was we could not run it in <50 minutes
<enebo>
but cloudbees has no restriction
<enebo>
but if we disabled it I don’t know why
<enebo>
heh
<bbrowning>
last build was nov 2014
<enebo>
So -0 should only be setting global $/
<bbrowning>
yeah, but -0 also ends up setting runtime.getDefaultSeparator() which puts looks at
<enebo>
but I see ‘if (processLineEndings) newBody.add(new GlobalAsgnNode(pos, "$\\", new GlobalVarNode(pos, "$/")));'
<bbrowning>
puts methods do "RubyString separator = (RubyString) runtime.getGlobalVariables().getDefaultSeparator();"
<bbrowning>
and $/ gets set to "globals.setDefaultSeparator(defaultRS);"
<bbrowning>
thus, -0 ends up controlling the puts separator
<enebo>
bbrowning: oh heh…seems like a bug in puts then
<bbrowning>
this is why I was looking for rubyspecs for this stuff :)
<enebo>
-0 is extremely esoteric for JRuby since no one uses us for quick oneliners
<bbrowning>
yeah
<enebo>
I am guessing RS in or out are very uncommon for non-oneliners too
<enebo>
which is why we probably don’t see or notice these bugs without ci coverage
<enebo>
no one tries to use them and we don’t ci all specs/tests/ which require CLI invocations because we start so slow
<enebo>
only difference is your snippet is setting the $/ explicitly
<enebo>
but I can see in default RS handler it is return “,”
<enebo>
which only makes me wonder if somehow the “,” returns is not matching due to encoding difference
<enebo>
but that makes no sense either since it is ascii
<enebo>
unless I am getting utf-16le bytes or something
<bbrowning>
hmm I still think the problem is just "-0" vs setting $/ explicitly
<bbrowning>
specifically, I think the problem is OpenFile#getlineFast
<bbrowning>
or maybe I'm wrong - I dunno. the IR does look the same to me as well
<bbrowning>
and I checked the encoding of RS when set via $/ vs -0 and it was UTF-8 for me in both cases
<enebo>
I will manually add the same assignment in the AST version and see if it still fails
camlow325 has quit [Ping timeout: 265 seconds]
<enebo>
HAHAH it works
camlow325 has joined #jruby
<bbrowning>
neat
<enebo>
but if I call getBytes() on the recordSeparator it breaks
<enebo>
so raw “,” is fine but the char made is not
iamjarvo has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<enebo>
no this does work but this is missing \n on the puts
<bbrowning>
1.7 was missing \n on the puts as well judging by that test
<bbrowning>
well the test looks for "1, ,2, ,3\n ," which is just weird
<enebo>
ah perhaps that is why there is the + ‘_’
<enebo>
or the + ‘ ‘
<enebo>
yeah but maybe it is correct
JRubyGithub has joined #jruby
<JRubyGithub>
[jruby] chrisseaton pushed 1 new commit to truffle-symbol-refactor: http://git.io/vIafZ
JRubyGithub has left #jruby [#jruby]
<JRubyGithub>
jruby/truffle-symbol-refactor a645073 Chris Seaton: [Truffle] Cache Symbol#to_proc
<enebo>
nope
<enebo>
so the test is wrong to boot
<bbrowning>
yeah test is wrong for MRI behavior, but does match what 1.7 did
<enebo>
bbrowning: possible it was 1.8ish
<bbrowning>
yeah perhaps so
<enebo>
well I may land this as unregressed but open issue that we are wrong
<enebo>
Something is really goofy
<enebo>
we have instrconfig.getrecordseparator and runtime.getRecordseparatorval and both seem to agree but it only works with the explicit $/ = foo line
<bbrowning>
it's magic
<enebo>
it makes no sense to me
<enebo>
All I can guess is there may be some logic which checks to see if it is the default sep and it does something else
<enebo>
or as you mention it gets confused in the bowels of gets and picks a newline path
<bbrowning>
there is one case where it does that, yeah
<bbrowning>
when it should be looking at runtime.getGlobalVariables().getDefaultSeparator()
<enebo>
well are you sure this is just not for newline case only?
<enebo>
DOH
<enebo>
yeah that is broken
<bbrowning>
but I don't see any failing tests or bugs around it so I wasn't sure if I was reading it right
<bbrowning>
or, like you said, just no one uses this
<enebo>
it is object equality check but we are changing that value to rs == defaultSet
<enebo>
bbrowning: sorry you did fine the bug
<enebo>
bbrowning: I can solve this a different way now that I read this
<enebo>
bbrowning: I could change recordSepVar when -0
<enebo>
bbrowning: but that is tough since runtime is not init’d yet
<bbrowning>
Don't go changing too much too close to release :)
<enebo>
bbrowning: So I think this should be a check against systemDefaultSeparator and make that
<enebo>
bbrowning: yeah I won’t do it now
<enebo>
bbrowning: this fix is ok for now but this logic is obviously wrong because -0 changes the default but this method unconditionally uses ‘\n'
<enebo>
bbrowning: newline is somewhat special in IO as it is plus it is only one byte in length
<enebo>
bbrowning: -0 is any octal value but I am unsure whether I could pick a multibyte octal?
<enebo>
bbrowning: if so then we cannot depend on -0 changing default and using a fast path
shellac has quit [Ping timeout: 258 seconds]
<bbrowning>
hmm yeah not sure about multibyte in any of this for the separator
<bbrowning>
only single bytes will work in any of this iirc
<bbrowning>
OpenFile uses int to represent the delimiter in appendline and getlineFast
<bbrowning>
and the slow path still uses appendline, which uses int to represent the delimiter
<enebo>
bbrowning: but int is enough for a code ponit and can be 4 bytes in size
<bbrowning>
ahh right
colinsurprenant has quit [Quit: colinsurprenant]
<enebo>
bbrowning: so I think the code in IO is right except it wants to know specifically if it the the default \n sep
<enebo>
so it maybe should be testing against that
<enebo>
if we knew we were changing to other 1 byte RS perhaps it would work (don’t know if it depens on \n itself for anything lower)
<enebo>
but that would only be an optimization
bb010g has quit [Quit: Connection closed for inactivity]
JRubyGithub has joined #jruby
<JRubyGithub>
[jruby] enebo pushed 1 new commit to master: http://git.io/vIamA
<JRubyGithub>
jruby/master b96813f Thomas E. Enebo: Unregress test_command_line_switches#default_dash_0_split_records
JRubyGithub has left #jruby [#jruby]
JRubyGithub has joined #jruby
<JRubyGithub>
[jruby] enebo opened issue #3026: The -0 option or any other initialization of InstanceConfig.getDefaultSeparator breaks gets http://git.io/vIasw
JRubyGithub has left #jruby [#jruby]
JRubyGithub has joined #jruby
<JRubyGithub>
[jruby] enebo opened issue #3027: test_command_line_switches#default_dash_0_split_records http://git.io/vIaGt
JRubyGithub has left #jruby [#jruby]
<enebo>
bbrowning: Thanks for your help again. The confusion has cleared
<bbrowning>
enebo: sure - thanks for fixing that test :)
bbrowning is now known as bbrowning_away
imperator has quit [Quit: This computer has gone to sleep]
<subbu>
that could be. I think there was one test that I rewrote long back that expected a certain GC timeline.
* subbu
is in a meeting
<enebo>
subbu: np. I am commenting this out with an issue
<bbrowning>
mjc_: in the case of your test, most of the time looks to be spent just opening and closing the file over and over again
shellac has joined #jruby
<mjc_>
yeah I think that is our problem too, trying to figure out how to do it in a threadsafe way in our app
JRubyGithub has joined #jruby
<JRubyGithub>
[jruby] chrisseaton pushed 4 new commits to truffle-symbol-refactor: http://git.io/vIaFU
<JRubyGithub>
jruby/truffle-symbol-refactor c2f0785 Chris Seaton: [Truffle] Switch RubySymbol fields to OM.
<JRubyGithub>
jruby/truffle-symbol-refactor b58bbf9 Chris Seaton: [Truffle] bytes to byteList in Symbol for consistency.
<JRubyGithub>
jruby/truffle-symbol-refactor 2eae47d Chris Seaton: [Truffle] Remove most usages of RubySymbol.
JRubyGithub has left #jruby [#jruby]
JRubyGithub has joined #jruby
<JRubyGithub>
[jruby] enebo opened issue #3028: IR temps holding references to ruby object in test/jruby/test_object_space#test_finalization http://git.io/vIabN
JRubyGithub has left #jruby [#jruby]
JRubyGithub has joined #jruby
<JRubyGithub>
[jruby] enebo pushed 1 new commit to master: http://git.io/vIaN8
<JRubyGithub>
jruby/master 9f76735 Thomas E. Enebo: Fix new id values for true,false,nil. Workaround #3028 so we can test the behavior of finalization in ObjectSpace
JRubyGithub has left #jruby [#jruby]
<enebo>
subbu: bbrowning: My trivial attempt to show it is temps holding values holds water
<bbrowning>
yep
<enebo>
bbrowning: but this should green up slow ci now
<mkristian>
enebo, so far so good, reached already jruby-core
djbkd has quit [Remote host closed the connection]
<enebo>
mkristian: is it even remotely possible me using a fresh local clone could be responsible? cd .. && rm -rf release && git clone jruby release && cd release
donV has joined #jruby
yfeldblum has joined #jruby
<enebo>
it did not work from my dev directory either
<mkristian>
enebo, let me try this as well since I need to switch away from orcale jdk since it does not do the javadocs with it
shellac has quit [Quit: Computer has gone to sleep.]
x1337807x has quit [Quit: My MacBook Pro has gone to sleep. ZZZzzz…]
nateberkope has quit [Quit: Leaving...]
<enebo>
mkristian: Either no login credentials were sent with the request, or login credentials which are invalid were sent. Checking the "authorization and authentication" system feed in the Nexus UI can help narrow this down. If credentials were sent there will be an entry in the feed.
<enebo>
mkristian: you know what that system feed is?
<mkristian>
no - but I am logged on to sonatype
<enebo>
mkristian: yeah me too
<mkristian>
enebo, usual question - which maven version are you using ? 3.3.3 ?
<enebo>
3.3.1
<enebo>
I can update
JRubyGithub has joined #jruby
<JRubyGithub>
[jruby] chrisseaton pushed 4 new commits to truffle-symbol-refactor: http://git.io/vIV2a
<JRubyGithub>
jruby/truffle-symbol-refactor 8af4e38 Chris Seaton: [Truffle] Simplify Symbol.all_symbols
<JRubyGithub>
jruby/truffle-symbol-refactor 5b0d8f8 Chris Seaton: Loosen the specs for Symbol#all_symbols increasing in size.
<JRubyGithub>
jruby/truffle-symbol-refactor 2cb7716 Chris Seaton: [Truffle] Sometimes we need to handle Rubinius undefined as our missing.
JRubyGithub has left #jruby [#jruby]
<enebo>
mkristian: same issue with 3.3.3
djbkd has joined #jruby
<mkristian>
hmm.
<mkristian>
on sonatype I have the same kind of account as you do. I did see this kind of error before but can not remember what solved it.
<enebo>
mkristian: so I can see the stream if I use Nexus Pro+
<enebo>
mkristian: so I am guess this means both client and server and we are using Nexus Pro
<enebo>
mkristian: maybe I will open an issue
<mkristian>
enebo, ok - I am sure oss.sonatype.org uses the lasted and most pro version of nexus. the client here is maven.
<enebo>
mkristian: no doubt but I the support doc claims I can look through the nexus ui to see failed auths
<mkristian>
yes an issue is one thing. and a last trial with maven-3.2.5 after running mvn install -Prelease with 3.3.x
<enebo>
mkristian: but it is working for you?
<enebo>
mkristian: I saw jruby-parent only
deobalds has quit [Ping timeout: 264 seconds]
<enebo>
mkristian: but maybe you interrupted the deploy
<mkristian>
javadocs failed with jdk8
donV has quit [Quit: donV]
<enebo>
ah yeah
<mkristian>
and now I started with a fresh clone
<enebo>
I use Java 7 for that reason
<mkristian>
I can not find any place in nexus UI where to see any logs
<mkristian>
enebo, for the rc1 - can we merge test-merge-1-7 before releasing. just got it green today
<enebo>
mkristian: what changed?
<mkristian>
well, merge current jruby-1_7 into master. big decimal and how to spawn jruby itself. and a couple foxes of file and dir operation using uri-like paths
sometimesfood has quit [Quit: bye]
<mkristian>
well, not important - just a question
<enebo>
mkristian: definitely not then. We cannot depend on ci alone to know if it ok
<mkristian>
ok
<enebo>
mkristian: more or less some bake time is required but we can land it after rc1.
<mkristian>
shall I continue the deploy ? reach truffle now
<enebo>
mkristian: this rc period is not likely to be without many changes
<enebo>
mkristian: no. I think it is just something wrong with my auth
<enebo>
I have zero idea which Jira project I should open this as
<enebo>
The Nexus one?
<mkristian>
enebo, no problem. let find you the link
<enebo>
mkristian: maybe I am using an older auth for sontatype?
<enebo>
mkristian: I have a username and password
<enebo>
mkristian: does ossshr use tokens?
<mkristian>
which project did you release yesterday ?
<enebo>
jnr-posix
<enebo>
which is a different group
<mkristian>
just want to compare the config to jruby
<enebo>
which config?
<mkristian>
I do have three username/password section. pom.xml == config how they name the server
shellac has joined #jruby
<enebo>
mkristian: three sections in ~/.m2/settings.xml?
robbyoconnor has joined #jruby
<mkristian>
sonatype-nexus-staging, ossrh and sonatype-nexus-snapshots
<mkristian>
they all have the same username/password I use to login on oss.sonatype.org
<enebo>
well I guess I only have the first and last one in ~/.m2/settings.xml
<enebo>
staging and snapshots also have the same password
<mkristian>
that are the one we use for jruby itself
<enebo>
mkristian: yeah and also must be fore com.github.jnr
<mkristian>
yes
<mkristian>
I can not find this issue tracker anymore :(
<enebo>
ok so I cannot release a snapshot of jnr project either
<mkristian>
this one of you maven days :)
<enebo>
mkristian: heh
<enebo>
mkristian: I will wipe out .m2
<enebo>
repository
<enebo>
perhaps I have some weird jar
<mkristian>
no this is not a solution only a trial
<enebo>
mkristian: ok I didn’t understand that
<mkristian>
never mind - wiping out the repo always means so much download after this
<enebo>
mkristian: could using ./mvnw changed some setting somewhere?
<mkristian>
you could move away .mvn and then run regular maven. then it is out of the box
camlow325 has quit [Remote host closed the connection]
<enebo>
mkristian: [ERROR] Child module /Users/enebo/work/release/test/pom.xml of /Users/enebo/work/release/pom.xml does not exist @
<enebo>
mkristian: so maybe I cannot do that but it ia immaterial since I cannot push a SNAPSHOT on a jnr-* project now
<mkristian>
just run mvn clean deploy -Poss-sonatype-release - for testing
<mkristian>
but right SNAPSHOT not working is as bad
<mkristian>
clean might not work - not sure though
<bbrowning>
mkristian: I suggested clean just because that's the fastest thing to dump out pom.xmls
<bbrowning>
basically acts as a no-op to just dump pom.xml and then you can use regular mvn to do the real stuff
<mkristian>
bbrowning, if it works - then ok. I used to use mvn validate - but ran in issue missing artifacts along the way when switching to a new version
emakris has quit [Quit: Leaving]
JRubyGithub has joined #jruby
<JRubyGithub>
[jruby] chrisseaton opened pull request #3029: Truffle symbol refactor (master...truffle-symbol-refactor) http://git.io/vIVbf
JRubyGithub has left #jruby [#jruby]
<chrisseaton>
Anyone interested in Truffle and worried about our warmup performance - example of how much better it is now here https://github.com/jruby/jruby/pull/3029
rsim has quit [Client Quit]
emakris has joined #jruby
iamjarvo has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
skade has joined #jruby
<lopex>
numbers!
pietr0 has joined #jruby
<enebo>
chrisseaton: warmup is not significant in any other impl but truffle so I not sure what to infer from it…that said those numbers look nice
tenderlove has joined #jruby
<mjc_>
chrisseaton: wow
<chrisseaton>
enebo: warmup used to be much much worse than that
<enebo>
chrisseaton: ah yeah that is part of the missing bit for me. I am guessing you can completely inline that proc generate block back to the site to get good overall prf
<enebo>
chrisseaton: how fast you can do it is what I cannot see
<enebo>
chrisseaton: very cool you guys are improving warmup
<chrisseaton>
Yeah the whole thing is inlined
<enebo>
hmm 9k beats 1.7 in this bench
<enebo>
I am not using indy
djbkd has quit [Remote host closed the connection]
<chrisseaton>
Yeah I didn't turn Indy on sorry
<enebo>
chrisseaton: it is fine
djbkd has joined #jruby
<enebo>
mkristian: ok totally worthless suport
<mkristian>
well
<enebo>
“Check to make sure the ID of your distributionManagement URL matches the ID in the server section of your settings.xml with the needed credentials, this is the most common cause of this issue.”
<mkristian>
let me see if I find an url like the curl example which ask for the credentials
<enebo>
mkristian: I wrote back pointing out my settings.xml has not changed since 2013 and you could deploy so I might get pushed to community support in ossrh to look at logs
<mkristian>
you could deploy yesterday but not anymore today without any changes on pom.xml and settings.xml
<enebo>
mkristian: I have not changed my settings.xml since 2013 and I did deploy yesterday
<enebo>
mkristian: but it was another project using the same auth
<enebo>
mkristian: I can log into their oss sontatype website but that obvious is different auth
<enebo>
mkristian: so either mvn is not forwarding my auth credentials
<enebo>
mvn help:effective-settings"
<enebo>
will that tell me?
<mkristian>
have a look if it is different from your settings.xml
<mkristian>
did you run mvn deploy with -X
<mkristian>
that might give you more hints what happens
<mkristian>
there I can post the pom.xml to oss.sonatype.org
bbrowning is now known as bbrowning_away
<mkristian>
does this work for you ?
<mkristian>
(I used -u mkristian btw)
<enebo>
mkristian: this it supports to be which pw?
<mkristian>
same password as I use for maven or the website
skade has quit [Ping timeout: 244 seconds]
<enebo>
mkristian: my website pw is completely different
<enebo>
mkristian: as is my username
<mkristian>
interesting. I have only one for all
<enebo>
<head><title>502 Bad Gateway</title></head>
skade has joined #jruby
<mkristian>
me too
<enebo>
so using mvn iuser/pw I got that
<enebo>
ok
<mkristian>
on the website
<mkristian>
first thought is me using TOR but also without it
<mkristian>
had an error about exceding capacity before.
<enebo>
mkristian: “I assume you are using a user token, since I ran into the same problem. We had some problems with the user tokens database and are currently restoring. Please stay tuned and wait for an update to test again.”
<enebo>
mkristian: so I do remember getting a token/pw combo when I got set up and I am guessing you are not using a token or they did not lose yours
<mkristian>
I never heard about those tokens before. I just have password
<enebo>
mkristian: well I have a username and a password in settings but they are random characters both