Closed 20kdc closed 5 years ago
I couldn't verify the macOS issues on my machine, let's see what CI says again
@fridtjof any luck?
Seems like there's a memleak on linux now. macOS looks fine though
...Does the CI build actually have debug symbols?
CFLAGS = -fsanitize=address,undefined
might be eliminating them.
It'd be very useful if AddressSanitizer could return precise debug information...
...actually, the debug info is added with:
CFLAGS += -Og -ggdb
so it should remain intact. New theory: The stacktraces just can't handle mod_dl
.
In this case, is there a way to get library base addresses out of dl
?
It's a more manual process but will help for tracking down the responsible module.
Well, we could log the addresses the libraries get loaded at with dlopen, that could help.
Before merging this, I'd like to match the macOS versions built against on travis and azure pipelines, to identify on which versions it fails to build.
However, Azure Pipelines doesn't have a build agent for macOS 10.12. This is the version that travis builds are failing on. This means we will have to either:
I can't assist with option 2, because my machine is running 10.14.
It fails on 10.13 too. https://travis-ci.org/shinyblink/sled/jobs/519066166
On that build, 10.12 built fine.
Awesome, @20kdc fixed the build issue. It was not macOS specific, it had to do with Make parallelity and k2link.
Rearchitecture of module system. I'm not sure this is really refined enough, but vifino would like it as a PR so it gets a PR. This also includes mod_farbherd since this was derived from that. In particular I'm concerned about what happens to the CI on Mac OS X. Advantages:
(EDIT : dammit markdown stop messing with my numbers)
Disadvantages: REALLY BIG ONES.
This fixes #66.