Time to retire USE_DATE_TIME_PRE_1_33_FACET_IO

[jeking3]

1.33.0 came out a long time ago. Time to retire the pre-1.33.0 code paths.

[eldiener]

I think you should query the developers mailing list regarding this potential removal.

[jeking3]

There was no feedback indicating this should not be done.

[eldiener]

I will remove it and update the 'develop' branch after I do.

[eldiener]

It looks to me that all sorts of tests are only being run when USE_DATE_TIME_PRE_1_33_FACET_IO is being defined. These includes tests for:

• date_time_wide_streaming
• date_time_pre_133_facet
• date_time_serialization

I do not see the same tests being run without USE_DATE_TIME_PRE_1_33_FACET_IO being defined. I feel that I am missing something in the bjam file for the tests but I can not see what it is.

This suggests to me that we can not remove the code which contains USE_DATE_TIME_PRE_1_33_FACET_IO.

[jeking3]

Hmm, interesting. Well I suggest we leave this here for now and revisit it once the trac backlog is drained a little more, or someone can pick it up.

It would seem that date_time_pre_133_facet would be unnecessary entirely based on the name alone.. not sure about the others.

[jeking3]

It's possible that serialization integration only happens pre-1.33 because of the streaming and traits introduced in 1.33... While enabling UBSAN builds I found I had to disable the serialization tests which only run in pre-1.33 compatibility mode. Might be more reason to remove that legacy code...

[eldiener]

If serialization needs the pre-1.33 code I do not see how we can remove that code without saying that serialization does not work for iostreams.

[eldiener]

Sorry I meant to say:

If serialization needs the pre-1.33 code I do not see how we can remove that code without saying that serialization does not work for date_time.

[jeking3]

True, will come back to it eventually and see.

[JeffGarland]

So I'm planning to remove support for this in boost 1.73 -- the time has seriously come. I've made an initial commit that removes the macros the force this mode for long dropped compilers. I posted a warning on both the mailing lists: crickets.

https://lists.boost.org/boost-users/2020/03/90303.php
https://lists.boost.org/Archives/boost/2020/03/248334.php

I have a 'dt_nolib' branch which I'm going to put onto develop shortly -- removing the library build. That changes the testing quite dramatically and as part of that I've dropped all the pre-1.33 tests. The code is not removed, yet -- but that will be soon.

[JeffGarland]

This has been merged to master for release in 1.73