It's the same behaviour as 7.8.116 though.

That doesn't make it any better, though.

Not does it make it any worse, compared to the current situation. I'm just saying this does not need to block a new release.

Why is that bad? It seems a _lot_ better than this:

(string)0.000000000000000000001;

(1) Result: "0.000000"

It's a lot better than something that translates to 0, in my opinion. At least (float)(string)x gives back approximately the same result now, as well as the string not being really really long for large numbers.

I agree, I just think that actual representation of the number is better, and should have a higher priority than a decimal dot not to misinterpret it as an integer.

Adding ".0" to the number if it's missing should be easy enough, so there isn't any reason not to have both, is there?

That doesn't seem correct,

sprintf("%.15g",1e100);

(1) Result: "1e+100"

shouldn't be 1e+100.0. (I can't test (string)"1e100" due to the fact it has too high precision right now.)

After some consideration and consulting I've decided it doesn't. We keep postponing releases for non-regressions which ultimately gives us these extremely long beta periods.

7.8.316 is going out as a release shortly. The next release will be in August.

yeah, but that is some work that needs testing. if the next release is aimed for next month then that's early enough (if someone picks it up and gets it to work by then.)

greetings, martin.

• I assume "some distributions" can be instanciated to "Gentoo"?

Well, Nettle 2.x is official GNU software since a week or two back, so I assume others will follow shortly.

7.8 is released, though.

What changes? The changes between Nettle 1.15 and 2.0 are minimal (but cause compilation failures).

Probably because the API changes were incompatible.

The main incompatible change is that a few typedefs went from function pointers to functions, which is trivial to detect in configure if you know about it...

I haven't talked about switching out the bundle, what I'm talking about is supporting using a pre-installed Nettle 2.0.

I mean n.m releases, not n.m.p.

I think there will be a 7.10 before 8.0.

Provided the multi-cpu project actually gets underway, it will probably start with a fairly big refactoring of the header files. The main reason is that a partly new API is necessary to track pointer assignments and to allow for the required changes in refcounting, but of course we'll probably take the opportunity to clean up the macro mess and lots of other old cruft. The plan is to do this work in 7.9, i.e. it will be part of 7.10.

When that is done, the actual multi-cpu development can start, without requiring further API changes in C modules etc. That branch would then become 8.0.

There are several reasons for this split: 1) Allow for gradual migration of C modules - in 7.9/7.10 they can work with both the old and the new API. 2) Split up the instability - both the refactoring and the multi-cpu changes will introduce bugs, and it's good to be able to separate them a bit. 3) To insulate other development from the multi-cpu mess better. 4) To be able to compare performance between the multi-cpu branch and the old one with minimal code differences.

I'd have a serious issue with that decision if I had to use a dist, but my own interests don't get hampered very much since we at Roxen build from cvs anyway.

If you had to use a dist, then you would not have wanted a dist to be made? I think making the dist was the correct decision. We can make a new dist in a few weeks.

o The regression from 7.6 is pretty bad. It's therefore fairly important that a fix gets out before 7.8 use gets even more spread.

This only says that a fix should be released soon. The options now was either

A) Release 7.8.316 now, and 7.8.X later with the fix B) Don't release 7.8.316, and 7.8.X later with the fix

Not releasing 7.8.316 does not make 7.8.X ready for release any sooner.

o The fix was already committed (with testsuite tests) well in advance of the final release date. 7.8.324 would have contained it and still be without the Nettle 2.0 stuff. It could have been released instead.

No. 7.8.324 has not undergone beta testing. When 7.8.324 has been beta-tested, it can be released. Releasing 7.8.316 does not limit the possibility to do so.

o The release has been held up for much more petty stuff than this.

Which is bad. Two wrongs don't make a right.

No, a 7.8.324 release around July 13:th was not possible, due to releng (ZinO) going away to Dalarna. I'll let him answer on his requirements for a beta-test phase when he gets back.