[PATCH] D90714: [clang]Fix length threshold for MicrosoftMangle md5 hash

[David Blaikie via Phabricator via cfe-commits]

dblaikie added a comment.

In D90714#2375911 <https://reviews.llvm.org/D90714#2375911>, @mibintc wrote:

> In D90714#2374913 <https://reviews.llvm.org/D90714#2374913>, @dblaikie wrote:
>
>> Since the same code is used to mangle all these things, probably just test one of them?
>>
>> Could use macros to stamp out longer names without having to write them out manually?
>
> Not sure what technique to use to stamp out longer names, I tried using token pasting but that doesn't work because the tokens aren't macro expanded before pasting?
> #define X50 xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
> #define X45 xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
> #define X100 X50 ## X50  // X100 is X50X50

Yeah, there's a quirk of the preprocessor that means you need an extra level of indirection, like this:

  #define C(X) X ## X
  // X2: X * 2
  #define X2(X) C(X)
  // X8: X * 2^3 (8)
  #define X8(X) X2(X2(X2(X)))
  // X4096: X * 8^4 (4096)
  #define X4096(X) X8(X8(X8(X8(X))))
  void X4096(x)() {
  }

https://godbolt.org/z/r6Evae

>   I like the test source I added because it came from a program seen in the wild, and it demonstrates that reasonable length identifiers can generate enormous mangled names.

I think using a very plain name (like "lots of 'x'") helps clarify the test case is only about length - not about any other property of the name. A comment describing how templates (& especially template default parameters) can lead to long names could be good, though.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D90714/new/

https://reviews.llvm.org/D90714