Conversation
|
Fixes #501 |
|
Thanks @teeks99 , now I only need to analyze and fix the fallout of the update. Very likely some configuration mistake or tool (faber) limitation with the new compiler versions... |
|
@teeks99, in fact, a bunch of those failures are caused by the link step trying to use a static Python library. As the Python versions are pre-packaged into those docker images you provide, can you check whether the Python installation comes with shared (Python) libraries ? |
|
Where are you seeing the shared python library errors? I'm doing my own builds for python2, so I'm betting I need to add to that. It looks like all the C++98 tests are failing due to other boost modules updating. I think the python 3 failures are in the test_upcast test, here? |
|
It seems that Boost.Function requires C++11 since 1.85 (boostorg/function@f86ddb0), and upcast fails since python is now 3.12 (#479). |
|
I noticed the Python archive being used in https://github.com/boostorg/python/actions/runs/18892704105/job/53923684026?pr=502 for example. Again, it's possible that my way of linking is wrong and the archive being picked up when the DSO should be used. I need to investigate... |
|
I've just pushed a new image of clang, with a python 2 version that added I'll try to get gcc's image to have the same flag in the morning. Right now something is up with ftp.gnu.org where I get the gcc source from, so holding on that. |
|
:sigh: that didn't seem to work, I'll have to poke around more at another time |
|
I've made another change to the clang image, running a I'm still seeing infrastructure problems in the gcc repos, so that one isn't available yet. I'll update when it has the same fix. |
|
I finally got a decent gcc-15 build, that has been pushed as well, it looks like everything (except the C++98 jobs) is getting past build and into test. |
|
I've integrated this into #505, where I've also addressed all the fallout of the upgrade. |
3 participants
No description provided.