Thoughts on versioning

Dukc ajieskola at gmail.com
Wed Oct 27 18:01:47 UTC 2021


On Tuesday, 26 October 2021 at 01:19:29 UTC, Andrei Alexandrescu 
wrote:
> Versioning Phobos would free us from maintaining backward 
> compatibility with a variety of decisions that did not 
> withstand the test of time

I have a relatively simple LTS scheme in mind:

- We agree on a scheme for long-term support Phobos versions. For 
the sake of example, let's say we agree on 2.096 and every sixth 
minor version afterwards (2.102, 2.108 and so on).

- Any released DMD is required to compile all LTS Phobos versions 
older than the DMD. LTS Phobos versions may be patched to satisfy 
that requirement, so no need to keep around any DMD bugs or old 
features Phobos depends on. However, the LTS Phobos versions do 
not aim to support using newer language features. Phobos bug 
fixes may but are not required to be backported.

- DRuntime always uses the same version as DMD, and thus must 
also be compatible with older Phobos LTS versions.

- We change the language to allow relative imports, and convert 
Phobos to use them. This is needed to allow using two or more 
different versions of Phobos side-by-side.

- After these are in place, we may proceed with relatively 
aggressive changes in Phobos.


These do not mean that big a maintenance burden, because the LTS 
versions are only required to be kept compiling - only a fraction 
of the effort that would be needed to backport the bug fixes or 
new features. Also the LTS Phobos would act as a good real-world 
regression test for DMD and DRuntime.

However, this plan requires that Phobos must stop calling 
DRuntime internals directly - otherwise it's going to cause lots 
of trouble with this scheme. I don't think that is a problem 
though - Phobos is supposed to be just another library. Surely it 
does not need special privileges to do its job? If it needs, 
DRuntime needs a redesign.


More information about the Digitalmars-d mailing list