Fix appender code that might not initialize memory that might point at huge allocations

[schveiguy]

Not going to reproduce all the junk I did in the bug report. See the report.

[dlang-bot]

Thanks for your pull request, @schveiguy!

Bugzilla references

Auto-close	Bugzilla	Severity	Description
✓	24856	normal	std.array.Appender.ensureAddable can create stale memory references

Testing this PR locally

If you don't have a local development environment setup, you can use Digger to test this PR:

dub run digger -- build "stable + phobos#9084"

[schveiguy]

switching to draft, need to add a test.

[schveiguy]

well, I guess there isn't any sane testing harnesses I can incorporate here, and I don't want to add one for this. So it will have to go in with existing tests.

[jmdavis]

well, I guess there isn't any sane testing harnesses I can incorporate here, and I don't want to add one for this. So it will have to go in with existing tests.

Wouldn't a test to make sure that the elements were zeroed out suffice?

[LightBender]

I'm inclined to agree with @jmdavis. Given the fragile nature of the proof-test and that it requires a whole-program testing framework that does not currently exist in Phobos, I would think that a test that ensures that the desired "correct state" is achieved should be sufficient for our purposes. Such a test would prevent regressions which is mostly what we're after.

[schveiguy]

I can try and add a unittest that attempts to cause the problem to happen. Let's see if I can get it to fail with the original code though.

[schveiguy]

OK, a test is added. I added a printout for the log in case the test isn't properly run. FWIW, when I tested locally without the fix, the test failed.

[thewilsonator]

otherwise looks good