[Swan-dev] fyi, we've run out of impair+debug lset_t bits
D. Hugh Redelmeier
hugh at mimosa.com
Thu Sep 20 04:54:03 UTC 2018
| From: Andrew Cagney <andrew.cagney at gmail.com>
| On Thu, 6 Sep 2018 at 11:53, Andrew Cagney <andrew.cagney at gmail.com> wrote:
Sorry for not replying sooner.
| > Short term we can get some extra bits by splitting debug and impair so
| > that they each have their own lset_t.
That seems pretty clean and non-controversial.
The original idea was that you could could have code selected by any
of several debug options, with the test being a single making
operation.
Then when impairments were added, it seemed easy to graft them onto
that mechanism. It had the right dataflow.
I don't think that a test for any of several impairment settings is in
our code.
(Only a few places has more than one debug flag been tested. I think
that we just haven't tuned our debugging code, but we should. Its a
verbose mess.)
| Turns out I need a lot more bits so ...
Interesting.
| > However, long term we'll need to come up with something different:
| > expanding lset_t somehow, or even using a new structure.
|
| I grafted on a second mechanism that sends pluto a pair of integers
| interpreted as WHAT:HOW (for instance, --impair ke-payload:omit) so
| the lset_t limitation is removed. Old code can continue to use the
| old flags.
|
| Andrew
|
| PS: I'm tempted to make the HOW a float so 'times' can also be sent
| over, later ...
PLEASE: NO TO FLOATS. They are the wrong abstraction.
Time can be handled as scaled integers.
More information about the Swan-dev
mailing list