| grayeul | I don't see a line in the deub output showing why it is writing the: 4C8907699141DCFF1B0B1A3B6A4D49D076402168.repo-id file. But I can see that it tries to load that. | 00:04 | |
| That's the (at least current) conflicting file... | |||
| timo | I don't remember exactly what .repo-id is responsible for. I think it points from a precompiled file to the repository it's made from? | 00:06 | |
| does it have different content in both rpms? | 00:07 | ||
| grayeul | I'll check | 00:15 | |
| different content.. but the same exact path | 00:16 | ||
| share/perl6/vendor/.precomp/CD2858E4B48D26E12AB7903CDD61A7201262E08D/4C/4C8907699141DCFF1B0B1A3B6A4D49D076402168.repo-id | 00:17 | ||
| content in one case is: 776B03BE312A46A163C0E2B6E0EBBF7C73127855 and in another: 759DBDB008BCCBE3E22CEB7BA29FB20169AA7CA3 | 00:19 | ||
| timo | so one of the rpms should have a folder with that name in it i think | 00:20 | |
| grayeul | that's what I would have expected :) | 00:21 | |
| I am looking on one system, where I installed *one* of these rpms -- happens to be file-directory-tree in this case. | |||
| But I did both a 'find' and a 'grep' through the whole /opt/rakudo tree (where I'm putting things) -- and don't see any files/dirs like that -- and the only place that content shows up is in that rep-id file | 00:22 | ||
| I might try killing that file to see what happens -- since it doesn't *seem* to point to anything -- what could go wrong? :D | 00:25 | ||
| timo | in any case, it's safe (but not efficient) to not put precompiled files into the packages | 00:58 | |
|
02:58
rakkable left
02:59
rakkable joined
13:24
rakkable left,
rakkable joined
|
|||
| grayeul | It looks like I can remove the .precomp/**.repo-id files that are created (that would appear to have conflicting content in different modules) and at least things 'work' This might be ok for now. | 13:25 | |
| However, I'm still having trouble with one module: Hash::Merge. When I try to build it within an 'rpmbuild' session (to package as an rpm) I get the following: gist.github.com/grayeul/c13cfdd6d7...b9294690d6 | 13:26 | ||
| lizmat | which Linux distrp is this on? | 13:27 | |
| *distro | |||
| grayeul | I'm trying to build for Rocky-9 | 13:28 | |
| Without all the debug on, the basic error reported is: | |||
| lang-call cannot invoke object of type 'VMNull' belonging to no language | 13:29 | ||
| lizmat | fwiw, I cannot repro this on MacOS (on Mac sililcon) | ||
| I've seen reports of exactly that error on Alpine | |||
| grayeul | I *think* the error relates to trying to 'use Hash::Merge' in the process of installing the module -- i.e in the Augment.rakumod | 13:30 | |
| I think it only happens within the special chroot being used for an rpmbuild... i.e. I can do the 'install' normally into the currently active raku areas | 13:31 | ||
| lizmat | interesting... | ||
| grayeul | timo: had pointed me at some .spec files used for opensuse to compare how they do it (also building rpms) -- but I couldn't find one for Hash::Merge... | 13:32 | |
| I'd still like to understand why building different modules (independently) seemed to produce (in some cases) the exact same .repo-id file name, but with different content. In that case, the rpms themselves build fine, but they can't be installed together because they are each trying to install conflicting files. | 13:33 | ||
| timo | which of the core files is /opt/rakudo/share/perl6/core/sources/4C8907699141DCFF1B0B1A3B6A4D49D076402168 ? | 13:34 | |
| grayeul | I seem to have worked around that (not sure of the consequences) by removing the .repo-id files in the vendor/.precomp area that are created.... not sure of the significance of /.precomp vs /precomp | ||
| time: how can I tell? Inspect contents? | |||
| timo | oh, .precomp? that's not supposed to go in vendor/ at all | ||
| you might get that if you use RAKULIB=/bla/bla/bla with that folder | |||
| grayeul | I did use that -- because I thought the problem was related to not finding modules that were just installed.... | 13:35 | |
| timo | do you make sure to set RAKULIB to something starting with inst# ? | 13:36 | |
| grayeul | Contents suggested the 4C8 file is: class CompUnit::Repository::Staging is CompUnit::Repository::Installation | ||
| no -- I just gave a dir path... I can try adding the inst# prefix -- I see that in the reported repository output. | |||
| timo | yeah that might be the problem | 13:37 | |
| grayeul | I haven't checked on the precomp question... but that did not fix my Hash::Merge build issue -- same error | 13:38 | |
| timo | let's tackle one thing at a time and see if that makes the "different rpms conflict with each other" issue go away for the modules that already built successfully | 13:39 | |
| grayeul | ok | ||
| it does look like that gets rid of the /.precomp files -- that's good! | 13:42 | ||
|
14:02
sp1983 joined
14:08
sp1983 left
|
|||
| grayeul | and it looks like maybe I didn't need to set RAKUDOLIB anyway... it didn't solve my Hash::Merge problem, and the other modules seem OK without it. | 15:04 | |
| Note of interest: I'm still trying to track it down, but I seem to be able to build my raku-hash-merge rpm 'natively' without a problem -- the failure only occurs when using 'mock' (rpm-software-management.github.io/mock/) but it would be easier to track down the real problem, if I knew why the raku install-mod is failing. | 15:06 | ||
| Geth | rakudo/main: 92ead2bdff | (Elizabeth Mattijsen)++ (committed using GitHub Web editor) | src/core.c/Grammar.rakumod Add some sanity checking on Grammar.parse (#6088) irclogs.raku.org/raku/2026-03-14.html#14:22 noted an LTA error message for: $ raku -e 'grammar { method TOP { <.asdfwqef> } }.new.parse("asdf")' P6opaque: no such attribute '$!pos' on type Match in a Str when ... (8 more lines) |
16:23 | |
| rakudo/main: 9dfe3e7302 | (Elizabeth Mattijsen)++ (committed using GitHub Web editor) | src/core.c/Mu.rakumod Check for literal value in signature with .return(-rw) (#6086) github.com/rakudo/rakudo/issues/6003 pointed out that all return typechecks were bypassed if a literal return value was specified in the signature **and** .return or .return-rw were used. $ raku -e 'sub a(--> 42) { "foo".return }; dd a' ... (21 more lines) |
16:24 | ||
| rakudo/main: d04cb70ebc | (Elizabeth Mattijsen)++ (committed using GitHub Web editor) | src/core.c/Rakudo/Iterator.rakumod Make .comb accept Cool values for limit (#6083) In response to github.com/rakudo/rakudo/issues/6030 So "foobar".comb(/\w/, 2.0) will produce ("f","o").Seq |
|||
| rakudo/main: 03f90b6cd0 | (Elizabeth Mattijsen)++ (committed using GitHub Web editor) | src/core.c/CompUnit/Repository/FileSystem.rakumod Only look at .pm for code in 6.c/d with -I (#6081) Exclude the .pm extension if running in v6.e or higher in CURFS |
16:42 | ||
| rakudo/main: 4cbb955334 | (Elizabeth Mattijsen)++ (committed using GitHub Web editor) | src/core.c/Argfiles.rakumod Add support for %*SUB-MAIN-OPTS<dash-as-STDIN> (#6078) If set, will *always* assume that a bare "-" on the command line will be interpreted as STDIN, regardless of language version. This feature to be added in light of the general deprecation of "-" as an indication for STDIN / STDOUT. However, many command ... (5 more lines) |
16:44 | ||
| rakudo/main: 245d07274c | (Elizabeth Mattijsen)++ (committed using GitHub Web editor) | src/core.e/Fixups.rakumod Make semantics of div / mod Int only in 6.e (#6070) In response and as a solution to: github.com/Raku/problem-solving/issues/326 github.com/Raku/problem-solving/issues/504 With having the behaviour of mod in there for multiple years, it ... (5 more lines) |
16:49 | ||
| rakudo/main: 577a04ae4d | (Elizabeth Mattijsen)++ (committed using GitHub Web editor) | 2 files Add :without-backtrace named argument to die() (#6076) * Add :without-backtrace named argument to die() Following up on: github.com/Raku/problem-solving/issues/59 that was discussed in the resolutions meeting of 21 Feb 2026. $ raku -e 'die "foo"' foo in block <unit> at -e line 1 |
17:02 | ||
|
18:04
finanalyst joined
|
|||
| finanalyst | [Coke]: ping | 18:04 | |
| .tell [Coke] there is a PR on the website requiring your /someone's review. It is a whitespace only problem | 18:05 | ||
| tellable6 | finanalyst, I'll pass your message to [Coke] | ||
| [Coke] | . | 18:11 | |
| commented on the PR - is there a rendered version of the change somewhere? | 18:12 | ||
|
18:24
david7832 joined
|
|||
| lizmat | m: dd (1,2,3) cmp (1,2,3) # all good | 18:47 | |
| camelia | Order::Same | ||
| lizmat | m: dd (lazy (1,2,3)) cmp (1,2,3) # meh | ||
| camelia | Order::Less | ||
|
20:32
finanalyst left
|
|||
| grayeul | For any that may care... looks like I solved my rpm-building problem with Hash::Merge by adding `--/precompile` to the install-dist.raku call. Still not sure what it was about the 'mock' environment that was messing that up. | 21:19 | |
| Geth | rakudo/lizmat-cmp: 60b62f3c94 | (Elizabeth Mattijsen)++ | src/core.c/Order.rakumod Add infix:<cmp>(Iterable:D, Iterable:D) In response to github.com/rakudo/rakudo/issues/6075 This will do a cmp on all the values that their iterators produce, and thus creating the same semantics for Seq cmp Seq or something with a Seq on either side, such as: dd (lazy (1,2,3)) cmp (1,2,3) which would produce `Less` instead of `Same` before this addition |
22:33 | |
| rakudo: lizmat++ created pull request #6102: Add infix:<cmp>(Iterable:D, Iterable:D) |
|||
|
23:40
david7832 left
|
|||