config-patches
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [PATCH] * config.sub: Deduplicate `basic_machine` by always allowing


From: John Ericson
Subject: Re: [PATCH] * config.sub: Deduplicate `basic_machine` by always allowing the vendor
Date: Tue, 14 Aug 2018 13:53:37 -0400
User-agent: Mozilla/5.0 (X11; Linux x86_64; rv:52.0) Gecko/20100101 Thunderbird/52.2.1

Hi Ben,

That wouldn't be much of a testsuite, now, would it? ;-)

Ben

Well, it's randomized property testing akin to [QuickCheck]. The point wouldn't be to test the accepted output, but just to ensure the input indeed is accepted. The manual tests are the only way to test the output, unless we use some sort of alternative spec like [LLVM]'s.

[QuickCheck]: https://en.wikipedia.org/wiki/QuickCheck

[LLVM]: http://llvm.org/doxygen/Triple_8cpp_source.html


----

Anyways so I started moving things but it felt arbitrary: I can't say in all cases where to draw the line in 1-component aliases and basic-machine short hands. Instead I have a different approach to do the deduplication without introducing lots of dead code. The second patch is the interesting one; it moves the big patterns to the bottom, as I hinted at earlier. This exposes some more overlapping patterns, changing behavior and triggering shellcheck (I don't know why the overlap of patterns matters to shellcheck, but it does). I think I need you to resolve these problems as you see fit. Then, with the basic machine checks so reordered, I can simply break the single case into two, and do the deduplication, without making any dead code or moving around anything.

The first patch is a few more tests. I'd write more, but I think triaging the fallout from patch 2 is more important, after which I can write tests with more confidence.

Sound OK?

John

Attachment: 0001-testsuite-config-sub.data-Add-legacy-test-cases.patch
Description: Text document

Attachment: 0002-config.sub-Move-the-big-patterns-to-the-bottom.patch
Description: Text document


reply via email to

[Prev in Thread] Current Thread [Next in Thread]