Set VERSION to 0.7.0-rc1

[ararslan]

The following need to be merged first:

• loading: make __precompile__ the default #26991
• delete replace!(pred, A, new) #27944
• RFC: rename UVError to IOError #28287
• fix #15722, support BigInts in precompiled modules and sysimg #28296
• Rename realmin/max -> floatmin/max #28302
• Deprecate nextpow2/prevpow2 #28304
• Bump Pkg #28309
• win10,tty: revert support for ANSI code passthrough #28314
• Implement sortslices, deprecate sortrows/sortcols #28332
• Housekeeping for rc1 #28334
• Add LLVM patch to fix #28360 #28362
• Add deprecations to the old iteration protocol shims #28365
• Add kw args to squeeze deprecation #28374

[StefanKarpinski]

must be lowercase

[tknopp]

What about #13099 ? It solves a real issue, which is a pain to work around in Gtk.jl. Also it seems that PR ist all set.

[JeffBezanson]

Yes, we could perhaps merge that now, but I would consider it non-breaking so it's not urgent to get it into any particular version.

[tknopp]

But it fixes a bug that persists since 3 years and leads to segfaults: JuliaGraphics/Gtk.jl#161. So its not like a feature request but a request to solve a bug which makes GUI programming with Gtk much harder than it is for other languages. I have code where the sigatom workaround cited in JuliaGraphics/Gtk.jl#161 does not work and in turn cannot proceed.

[JeffBezanson]

If everybody is allowed to block the release on a bug of their choosing, there will never be a release. I do want to merge #13099 though. Hopefully we'll be able to merge it by 1.0.x.

[tknopp]

sorry, I did not want to say that this is blocking the release just bringing to attention that there is a fix for a bug leading to a segfaults. From other segfault reports I had the impression that there is some interest getting these fixed prior to a release. But yes, no urgent need. Gtk.jl will require time for being updated to 1.0 anyway.

[ararslan]

Looks like @rfourquet added #27944, which isn't on the milestone nor marked for triage. Yea or nay?

[rfourquet]

Looks like @rfourquet added #27944, which isn't on the milestone nor marked for triage. Yea or nay?

Yes, with a question mark because it was rejected in triage but now Jeff re-opened it. It's a conservative Yea for me, but definitely not crucial (but should be decided before rc1 I think).

[JeffBezanson]

#27944 deletes functions that were only added in 0.7, so it's low risk and the concern is avoiding adding an API that we might not want to commit to.

#13099 is non-breaking but involves a complete redesign of the task code, so is quite risky. It seems to work well as far as any of us know, but we all know how that goes.

[ararslan]

#28310?

[maleadt]

I'd like to add #28291, which is a pretty small bugfix that makes it possible to @profile CUDAnative code. Review & approved, CI is green, so good to go but nonetheless putting it here for consideration.

[ararslan]

Somehow that checklist seems to keep getting longer... 😉

[StefanKarpinski]

It just got shorter—we're not doing #28193.

[ararslan]

HERE. WE. GO.