Re: [ANNOUNCE] Git v2.19.0-rc0

On Thu, Aug 23, 2018 at 04:30:10PM -0700, Jacob Keller wrote:

On Thu, Aug 23, 2018 at 9:30 AM Jeff King [off-list ref] wrote:

quoted

I think that audit isn't actually too bad (but definitely not something
we should do for v2.19). The cocci patch I showed earlier hits most of
them. It misses the negated ones (e.g., "if (oidcmp(...))"). I'm not
sure if there's a way to ask coccinelle only for oidcmp()
used in a boolean context.

You can look for explicitly "if (oidcmp(...))" though. I don't know if
you can catch *any* use which degrades to boolean outside of an if
statement, but I wouldn't expect there to be too many of those?

Yeah, that was my thought, too. And I've been trying this all afternoon
without success. Why doesn't this work:

  @@
  expression a, b;
  @@
  - if (oidcmp(a, b))
  + if (!oideq(a, b))

I get:

  Fatal error: exception Failure("minus: parse error: \n = File
  \"contrib/coccinelle/oideq.cocci\", line 21, column 0,  charpos =
  221\n    around = '', whole content = \n")

If I do:

  - if (oidcmp(a, b)) { ... }

that seems to please the parser for the minus line. But I cannot include
the "..." on the plus line. Clearly the "..." part should be context,
but I can't seem to find the right syntax.

FWIW, I do have patches adding hasheq() and converting all of the
!oidcmp() cases. I may resort to hand-investigating each of the negated
ones, but I really feel like I should be able to do better with
coccinelle.

-Peff