Fork me on GitHub
#clojure-dev
<
2019-10-25
>
andy.fingerhut21:10:44

@alexmiller I would like to do a core.rrb-vector release, say call it version 0.1.1 since it is a fairly small change since 0.1.0 release. If this is something I can trigger myself, happy to learn how. If not, please do so when you have the time. Not urgent.

seancorfield22:10:08

@andy.fingerhut When you login to the build server and go into that project, do you see a Maven release option in the left hand column?

andy.fingerhut22:10:38

I do now, thanks to Alex setting up an account for me on http://build.clojure.org. Things look to be progressing well.

andy.fingerhut22:10:51

Now if I can just find even one bug-free implementation of RRB vectors anywhere, out of the half dozen I have found so far. They all seem to be subtly wrong.

andy.fingerhut22:10:33

I have learned that while the Scala folks spearheaded the data structure, it isn't part of the standard Scala library yet, and the library developed in 2015 in Scala as part of a grad student's thesis prints error messages when running its included tests with extra debug checking enabled... Call me Ahab.

dpsutton22:10:05

godspeed ahab

seancorfield22:10:47

This talk reminded me to check the build server. I broke core.memoize 😐 Note to self: update is not portable across older versions of Clojure!

andy.fingerhut22:10:24

Yep, definitely a good idea to check the build server after your last commit before a release.

seancorfield22:10:45

And that's especially bad since I actually have a script that runs the tests on Clojure 1.6 up through master... I just didn't run it 😞

andy.fingerhut22:10:55

So easy to use features we take for granted now.

seancorfield22:10:52

Yeah, our code base at work wouldn't even run on 1.9 at this point. And next.jdbc was designed for 1.10 and up. I think core.cache and core.memoize support the widest range of versions right now...

alexmiller22:10:47

I can modify the min clojure version for that lib if needed

seancorfield22:10:27

No, the matrix did the right thing -- testing against 1.7 which showed the bug.

seancorfield22:10:36

When did update get added? 1.8? 1.9?

andy.fingerhut22:10:22

1.7 according to :added metadata

seancorfield22:10:34

I don't honestly expect anyone is really running 1.6 these days but I am supposed to test them against that... I mean, I wrote the script for it. I should actually use my own script 🙂

andy.fingerhut22:10:39

which I ran a program a few months back to verify is correct for almost all functions

andy.fingerhut22:10:48

(the only issues I found were with some that say they were added in a version 1.3 or earlier, but were actually added in a slightly different version 1.3 or earlier)

seancorfield22:10:30

Ah, I misread the build output. The 1.7 I saw was part of the Clojure plugin (1.7.1). I can't actually see where it is using 1.6 but that makes sense.

alexmiller22:10:55

the build is run with the clojure.version set in the pom.xml

alexmiller22:10:49

the matrix test is based on the settings in https://github.com/clojure/build.ci/blob/master/ci_data.clj - the default set is line 8-9, and then you can set a minimum out of that set on a per-project basis

seancorfield23:10:16

Thanks. Been a while since I needed to touch any of that. When did the matrix stop testing against 1.5.1? I just ran the core.cache and core.memoize tests and they still run on 1.5.1 (but not 1.4 because they require reduced).