Opening investigation into persistent build failure present in latest build
Seems to have first failed 2018-12-28 in build #1577
Last success was 2018-12-19 build #1553
The intervening commits are:
968ed91 2018-12-28 Nicolas Cellier - Generate VM code from VMMaker.oscog-nice.2501 — |
Side note, just found this command...
— |
In reply to this post by David T Lewis
Doing a recursive diff between 0ab0a27 and 968ed91 shows changes under only under src directories. Searching on "failed" in a diff of their respective build logs shows...
— |
In reply to this post by David T Lewis
Since I only regenerated VM sources from VMMaker, and that sources are shared with Squeak/Newspeak, and that Squeak/Newspeak ARM builds success, I convinced myself that it was yet another infrastructure problem with PharoVM. But if it's cogit.o that failed, then it would be interesting to inquire the diffs inside a #ifdef Pharo... Note that 0ab0a27 has [skip CI] — |
In reply to this post by David T Lewis
Diff of cogitARMv5.c between 0ab0a27 [GOOD] and 968ed91 [BAD]. — |
In reply to this post by David T Lewis
— |
In reply to this post by David T Lewis
Hi Nicolas, Just adding to what Ben wrote above, the error from the compiler is:
— |
In reply to this post by David T Lewis
Hmm, forget what I wrote above, That code appears to have been removed in later commits (f954a91). cogitXXXX.c generates LOTS of warnings, but appears to build. Sorry for the noise. — |
In reply to this post by David T Lewis
OK, I see it now, re-generating cogitARMv5.c adds the definition of unalignedLongAt. Eliot has already fixed that in VMMaker.oscog-eem.2502 so as to :
Either I revert, or I try to generate with 2502... — |
In reply to this post by David T Lewis
Already fixed — |
In reply to this post by David T Lewis
Closed #333. — |
In reply to this post by David T Lewis
Are you sure? https://travis-ci.org/OpenSmalltalk/opensmalltalk-vm/builds/474186755?utm_source=github_status&utm_medium=notification is Cog (HEAD) with just the ARM platform and SqueakFFIPrims moved from plugins.int to plugins.ext. The same error occurs in both squeak and pharo. So the problem appears to be something else than SqueakFFIPrims. But bed time... — |
In reply to this post by David T Lewis
Sorry, I believed you when you told that cogitXXX was building OK, and it build OK. Still, there is a FFI problem, look in the console:
This happens for Pharo build and in the Squeak build too.
bed time too... — |
In reply to this post by David T Lewis
Reopened #333. — |
Might this be as simple as the macro for the cpu type being wrong/tested wrong? IIRC we actually use ARMv5 witihin the cog so maybe that is causing confusion? tim -- tim Rowledge; [hidden email]; http://www.rowledge.org/tim Strange OpCodes: DO: Deadstart Operator |
Hi Tim, On Wed, 2 Jan 2019 at 02:24, tim Rowledge <[hidden email]> wrote: > > > Might this be as simple as the macro for the cpu type being wrong/tested wrong? IIRC we actually use ARMv5 witihin the cog so maybe that is causing confusion? Do you know which macro it should be? The options appear to be: __ARM_ARCH__ __arm32__ ARM32 I'll take a guess at ARM32 first. Cheers, Alistair |
In reply to this post by David T Lewis
So,
I would say, we can make mistakes, but it will be a lot lot easier to correct if we decompose the modifications! — |
In reply to this post by David T Lewis
Further investigation in #334 — |
In reply to this post by David T Lewis
Fixed in 1c6552d — |
In reply to this post by David T Lewis
Closed #333. — |
Free forum by Nabble | Edit this page |