CI run with minimum versions of dependencies #15136
Labels
low effort
There is not much implementation work to be done. The task is very easy or tiny.
low impact
Changes are not very noticeable or potential benefits are limited.
should have
We like the idea but it’s not important enough to be a part of the roadmap.
testing 🔨
Currently we only consistently test the code with one set of dependency versions on each platform. It's usually either the latest one of the one installed by default. If we use anything else, it's coincidental - when a different version happens to come with some non-standard CI image.
As pointed out in #15117 (comment), we should really have a CI run with the minimum supported versions to detect breakage. In most cases this breakage is harmless, because it only results in a failure to build, but there have been cases of semantic changes in Boost that resulted in the compiler producing different output (e.g. boostorg/multiprecision#129 or all the various changes to
boost::filesystem
that affect import processing). Nowadays these will often be caught by tests, but to catch them, have to run them in the first place.I think it would be enough to have such a run on Linux, as we support the widest range of versions, so it should catch vast majority of problems of this kind.
The text was updated successfully, but these errors were encountered: