[tz] [PROPOSED] Simplify zdump average-taking

John Hawkinson jhawk at alum.mit.edu
Thu Dec 1 09:27:56 UTC 2022

Paul Eggert <eggert at cs.ucla.edu> wrote on Thu,  1 Dec 2022
at 00:01:47 EST in <cc9a7cc5-ddbf-94dc-c5ad-faad18437abc at cs.ucla.edu>:

> This particular change's main point was to fix the comment, since we are moving to C99 or later.

The idea that moving to C99 means we should go out and remove C89-isms is troubling to me, because I don't think our test coverage is strong enough to ensure that these kinds of changes do not introduce problems. To me, moving to C99 means we don't go out of our way to preserve C89-compatibility, but not that we needlessly alter working code just because we can now use a C99 feature to simplify it.

(There is also an issue of burden on code reviewers.)

jhawk at alum.mit.edu
John Hawkinson

> The code simplification was a nicety that's independent from the C99
> change. The delta to the generated instructions was more a note to
> myself than anything important.
> I did run the usual regression tests, as I do with all code
> changes. You can run them too, by invoking "make check" or (my
> favorite on GNU/Linux) "make CFLAGS='$(GCC_DEBUG_FLAGS)' check". Of
> course the tests are not exhaustive.
> zdump is by far the biggest performance bottleneck in 'make check'
> so at least one user (namely me) would benefit from a faster zdump,
> not that this change affected zdump's performance significantly.

More information about the tz mailing list