Why do we discard (repeatedly) all warning filters?

[Andrey Novoseltsev]

At the moment any time a deprecation warning is issued all warning filters are thrown away:
https://github.com/sagemath/sage/blob/master/src/sage/misc/superseded.py#L138
which means that Python functionality for tuning warnings is completely useless. Why??? If the point is to change the default behaviour of not showing deprecation warnings, how about just adjusting that particular filter and only once when Sage starts?

[Johan S. R. Nielsen]

I asked a similar question recently:
https://groups.google.com/forum/#!topic/sage-devel/RxMNjVWBeI0

There seems to be some code outside sage/misc/superseded.py which
removes some warnings so they will be issued only once. I have no idea
how or where, or if it's just because I can't read the warning code in
superseded.py right...

I ended up monkey-patching the one type of warning I was interested in
at the time: http://trac.sagemath.org/ticket/20601

Best,
Johan

[Jeroen Demeyer]

On 2016-05-23 00:03, Andrey Novoseltsev wrote:
> Why???

I think it's safe to assume that there is a good reason, it's just the
way it was done historically.

[Jeroen Demeyer]

Sorry, I meant to say "I think it's safe to assume that there is *no*
good reason"...

[Andrey Novoseltsev]

Getting rid of resetwarnings in all places is ready for review at http://trac.sagemath.org/ticket/20667