rewording of CHANGES.txt info. Changed versionadded to 4.5.0#8
Conversation
This is not a hill I'm going to die on, but the version we're making the change in is 4.5; if there are micro versions, we promise not to make API changes in those, so I see no reason to distinguish 4.5.0 specifically. |
| dictionary keys is now preserved when generating the command line. | ||
| Previously these were sorted alphabecially: this may cause a change | ||
| which leads to a rebuild. | ||
| NOTE: If you use a dictionary to specify your CPPDEFINES, you may see unexpected build results |
There was a problem hiding this comment.
"unexpected" seems a bit strong when combined with "results". Unless the possibly changed order affects the actual build, which is really unlikely, it's a single rebuild (and silently, update of sconsign), but the result will be the same.
There was a problem hiding this comment.
If you update SCons to 4.5.0 and you didn't change any sources, and your project rebuilds. I'd call that unexpected.
This has long been the practice of SCons to warn about such, becuase we almost always will get questions about such if we don't note them.
There was a problem hiding this comment.
A rebuild may be unexpected, and this is warning about it. But "unexpected...results" suggests the build may not end up the same, which sounds worse than a rebuild to me. How about this:
NOTE: If you use a dictionary to specify all or part of CPPDEFINES, you may
see one-time rebuilds due to changed order of defines on the command
line. SCons now retains the order of addition of all entries to
CPPDEFINES. In previous versions a dictionary used for addition was
sorted alphabetically before the individual items were added.
There was a problem hiding this comment.
change to unexpected rebuild ?
There was a problem hiding this comment.
unexpected rebuild is more palatable than unexpected result....
The first release in the 4.5 series will be 4.5.0, and that's where this change is first introduced. |
There have never been versionadded's in docstrings before I started pushing them so there's no change to existing practice involved. Besides, mimicking what CPython does... and they have a ton of micro releases over the life of a release, unlike us. e.g. https://docs.python.org/3/library/datetime.html#datetime.date.fromisoformat |
|
Version strings should be consistant across all of SCons, IMHO. |
See description and:
There will be no version 4.5, these changes will be in version 4.5.0
(There may be a 4.5.1)
Also I think the CHANGES.txt changes are little easier to read and lead with the fact this is all about resolving outstanding issues with CPPDEFINES.
I think it's more useful to list the new test coverage, rather than the files they're changed/added in.