@disable TypeInfo.init #1766

aG0aep6G · 2017-02-17T22:04:46Z

As planned (one release behind schedule).

aG0aep6G · 2017-02-17T22:10:26Z

I'm not sure if I got the changelog entry right. The README.md didn't actually tell me to create a file, but that's the plan, right? When I build ../dlang.org/web/changelog/pending.html as instructed, my entry doesn't show up. The doc tester doesn't show it either. Maybe the tooling isn't there yet?

wilzbach · 2017-02-17T23:00:21Z

I'm not sure if I got the changelog entry right. The README.md didn't actually tell me to create a file, but that's the plan, right?

Yep - see Phobos for a couple of examples: https://github.com/dlang/phobos/tree/master/changelog

When I build ../dlang.org/web/changelog/pending.html as instructed, my entry doesn't show up.

Your problem is that the changelog entry doesn't end with .dd. It seems that the druntime changelog README is a bit outdated in comparison to Phobos: https://github.com/dlang/phobos/tree/master/changelog

For reference, that's how it looks once the file has the proper extension:

The doc tester doesn't show it either. Maybe the tooling isn't there yet?

Yep: dlang/dlang.org#1549

aG0aep6G · 2017-02-17T23:34:13Z

Blocked by dmd tests using TypeInfo.init - dlang/dmd#6547.

Your problem is that the changelog entry doesn't end with .dd.

Fixed, thanks.

aG0aep6G · 2017-02-18T00:39:31Z

Blocked by dmd tests using TypeInfo.init - dlang/dmd#6547.

That PR has been merged. auto-tester is turning green. So this should be good to go. I've squashed the fixup commit.

rainers · 2017-02-18T10:08:54Z

You should also update dmd not to use TypeInfo.init, as the travis build for dmd rebuilds itself, see for example https://travis-ci.org/dlang/dmd/jobs/202809672#L5264.

aG0aep6G · 2017-02-18T12:56:08Z

You should also update dmd not to use TypeInfo.init, as the travis build for dmd rebuilds itself, see for example https://travis-ci.org/dlang/dmd/jobs/202809672#L5264.

dlang/dmd#6550

aG0aep6G · 2017-02-18T13:26:23Z

Alright, we have a problem here. dlang/dmd#6550 fails because the test bots use relatively old releases to build dmd. Those don't have TypeInfo.initializer, so they fail.

I'm not sure, is there an older release that's deemed stable, which is supposed to be able to build all newer releases? If so, can we bump that to a newer one or make a point release for it? If not, can we just make the bots use newer versions?

wilzbach · 2017-02-18T14:23:58Z

I'm not sure, is there an older release that's deemed stable, which is supposed to be able to build all newer releases? If so, can we bump that to a newer one or make a point release for it? If not, can we just
make the bots use newer versions?

That would be 2.072.2. For Travis and CircleCi this is already used at the dmd repo (though not for druntime and phobos, but the bump is in the queue: dlang/phobos#5140, #1725)

However as the auto-tester (and GDC) still use an old frontend, you should definitely see this discussion http://forum.dlang.org/thread/FA142E6B-DD11-43F6-8EF9-EEC72AA1A7F2@klickverbot.at

aG0aep6G · 2017-02-18T16:44:23Z

That would be 2.072.2.

2.072 is not a long-term support version, is it? Anyway, 2.072 is fine, as TypeInfo.initializer has been added in 2.070.0. Trouble is with anything older than that.

However as the auto-tester (and GDC) still use an old frontend

The auto-tester and the doc tester are the real problem. If gdc is supposed to be able to build dmd, then it needs updating too. Question is how do I get TypeInfo.initializer into the testers (and maybe gdc)?

PetarKirov · 2017-02-21T07:33:49Z

@ibuclaw @jpf91 we would like to move forward with removing TypeInfo.init. Any chance of cherrypicking the druntime change from #1403 into GDC, so we can bump the minimum compiler version that is used to build DMD?

CC @MartinNowak

ibuclaw · 2017-02-23T16:33:56Z

GDC is not affected by this as far as I can see, the dmd builds are still passing on travisci.

aG0aep6G · 2017-03-13T14:22:35Z

dlang/dmd#6550 has been merged. Testers are green. Regarding dmd, this should be good to go.

gdc still uses the C frontend, so it should not be affected.

ldc will have to incorporate dlang/dmd#6550 to be be buildable by dmd 2.074. I've tried to cherry pick it, but the path to rmem.d is different in ldc and that makes things weird. @klickverbot: Please advise or take over.

dnadlinger · 2017-03-13T15:18:49Z

changelog/disable-TypeInfo.init.dd

@@ -0,0 +1,6 @@
+TypeInfo.init has been `@disable`d.


Perhaps write this out as init() to make clear that this is referring to the property function?

dnadlinger · 2017-03-13T15:27:14Z

changelog/disable-TypeInfo.init.dd

@@ -0,0 +1,6 @@
+TypeInfo.init has been `@disable`d.
+
+This is a step towards removal of `TypeInfo.init`, which is necessary to resolve


Maybe phrase this in a more direct and comprehensive way to give people who have not been following the process enough to make sense of the change? Something like:

TypeInfo.init() (deprecated since 2.0XX) collides with the built-in type property of the same name and will be removed in XXX. To get the default initializer of the described type, use TypeInfo.initializer instead.

(Okay, that isn't much of an improvement either, but I hope it gets my point across.)

dnadlinger

This is fine from an LDC perspective.

If we wanted to be cautious about reducing the number of future bootstrapping steps, we would make sure that jump straight past this change from the last C++ compiler, but we can always do this by pushing out appropriate LDC patch releases if necessary.

aG0aep6G · 2017-03-13T16:29:21Z

I've tweaked the changelog entry a bit, reacting to @klickverbot's comments, hopefully making things clearer. I don't really care about the exact words. Please feel free to just edit it, if you think it can be improved.

This is fine from an LDC perspective.

Alright. Looks like we're all set then.

MartinNowak · 2017-03-13T17:20:48Z

If we wanted to be cautious about reducing the number of future bootstrapping steps, we would make sure that jump straight past this change from the last C++ compiler

Bootstrapping will really only remain feasible via cross-compilation. It's just unrealistic to assume that we can rebuild every intermediate compiler in 10 years (not to mention the crazy amount of steps).

TypeInfo_Class.init is deprecated and removed from the runtime (since .init clashes with the type property of the same name). This uses `initializer()` method which is the new way of accessing the same thing in D2. dlang/druntime#1403 dlang/druntime#1766 dlang/druntime#1807 https://issues.dlang.org/show_bug.cgi?id=15037 https://issues.dlang.org/show_bug.cgi?id=12233

aG0aep6G mentioned this pull request Feb 17, 2017

TypeInfo.init -> .initializer dlang/dmd#6547

Merged

aG0aep6G force-pushed the disable-TypeInfo.init branch 2 times, most recently from 7815a1d to aca7193 Compare February 18, 2017 00:38

aG0aep6G mentioned this pull request Feb 18, 2017

TypeInfo.init -> .initializer (part 2) dlang/dmd#6550

Merged

wilzbach added this to the 2.074.0 milestone Mar 5, 2017

aG0aep6G force-pushed the disable-TypeInfo.init branch from aca7193 to 9b73c51 Compare March 13, 2017 11:33

@disable TypeInfo.init

c5bdd29

dnadlinger reviewed Mar 13, 2017

View reviewed changes

dnadlinger approved these changes Mar 13, 2017

View reviewed changes

aG0aep6G force-pushed the disable-TypeInfo.init branch 2 times, most recently from 88d380d to a4616e8 Compare March 13, 2017 16:02

changelog tweaks

47f9fe2

aG0aep6G force-pushed the disable-TypeInfo.init branch from a4616e8 to 47f9fe2 Compare March 13, 2017 16:10

MartinNowak merged commit ffe3a95 into dlang:master Mar 13, 2017

aG0aep6G deleted the disable-TypeInfo.init branch March 14, 2017 15:57

nemanja-boric-sociomantic mentioned this pull request Jan 19, 2018

Use TypeInfo_Class.initializer() in D2 builds to get the initializer sociomantic-tsunami/swarm#254

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

@disable TypeInfo.init #1766

@disable TypeInfo.init #1766

aG0aep6G commented Feb 17, 2017

aG0aep6G commented Feb 17, 2017

wilzbach commented Feb 17, 2017

aG0aep6G commented Feb 17, 2017

aG0aep6G commented Feb 18, 2017

rainers commented Feb 18, 2017

aG0aep6G commented Feb 18, 2017

aG0aep6G commented Feb 18, 2017

wilzbach commented Feb 18, 2017

aG0aep6G commented Feb 18, 2017

PetarKirov commented Feb 21, 2017

ibuclaw commented Feb 23, 2017

aG0aep6G commented Mar 13, 2017

dnadlinger Mar 13, 2017

dnadlinger Mar 13, 2017 •

edited

dnadlinger left a comment

aG0aep6G commented Mar 13, 2017

MartinNowak commented Mar 13, 2017

		@@ -0,0 +1,6 @@
		TypeInfo.init has been `@disable`d.

		This is a step towards removal of `TypeInfo.init`, which is necessary to resolve

@disable TypeInfo.init #1766

@disable TypeInfo.init #1766

Conversation

aG0aep6G commented Feb 17, 2017

aG0aep6G commented Feb 17, 2017

wilzbach commented Feb 17, 2017

aG0aep6G commented Feb 17, 2017

aG0aep6G commented Feb 18, 2017

rainers commented Feb 18, 2017

aG0aep6G commented Feb 18, 2017

aG0aep6G commented Feb 18, 2017

wilzbach commented Feb 18, 2017

aG0aep6G commented Feb 18, 2017

PetarKirov commented Feb 21, 2017

ibuclaw commented Feb 23, 2017

aG0aep6G commented Mar 13, 2017

dnadlinger Mar 13, 2017

Choose a reason for hiding this comment

dnadlinger Mar 13, 2017 • edited

Choose a reason for hiding this comment

dnadlinger left a comment

Choose a reason for hiding this comment

aG0aep6G commented Mar 13, 2017

MartinNowak commented Mar 13, 2017

dnadlinger Mar 13, 2017 •

edited