#10970: Built in MIN_VERSION macro support -------------------------------------+------------------------------------- Reporter: ezyang | Owner: Type: feature request | Status: new Priority: normal | Milestone: Component: Compiler | Version: 7.11 Resolution: | Keywords: Operating System: Unknown/Multiple | Architecture: | Unknown/Multiple Type of failure: None/Unknown | Test Case: Blocked By: | Blocking: Related Tickets: | Differential Rev(s): Phab:D1349 Wiki Page: | -------------------------------------+------------------------------------- Comment (by Ben Gamari ): In [changeset:"a9c93bdd8b027d6de09a3eada7721e7fd2d3e050/ghc" a9c93bdd/ghc]: {{{ #!CommitTicketReference repository="ghc" revision="a9c93bdd8b027d6de09a3eada7721e7fd2d3e050" Implement MIN_VERSION and VERSION macros natively in GHC. Test Plan: validate Reviewers: austin, thomie, bgamari Reviewed By: thomie Subscribers: thomie Differential Revision: https://phabricator.haskell.org/D1349 GHC Trac Issues: #10970 }}} -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/10970#comment:2 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler

#10970: Built in MIN_VERSION macro support -------------------------------------+------------------------------------- Reporter: ezyang | Owner: Type: feature request | Status: new Priority: normal | Milestone: Component: Compiler | Version: 7.11 Resolution: | Keywords: Operating System: Unknown/Multiple | Architecture: | Unknown/Multiple Type of failure: None/Unknown | Test Case: Blocked By: | Blocking: Related Tickets: | Differential Rev(s): Phab:D1349 Wiki Page: | -------------------------------------+------------------------------------- Comment (by ezyang): The big problem is that, if you do this, you effectively force GHC to dump the ENTIRE package database into a giant header file, which defines each of the macros. The bigger the database, the bigger the file. I'm not fundamentally opposed, but this was the reason this draft didn't do it that way. I think Duncan also has some ulterior motives for not enabling it by default, mostly to get people to use proper setup dependencies on their Cabal files. -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/10970#comment:4 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler

#10970: Built in MIN_VERSION macro support -------------------------------------+------------------------------------- Reporter: ezyang | Owner: Type: feature request | Status: new Priority: normal | Milestone: 8.0.1 Component: Compiler | Version: 7.11 Resolution: | Keywords: Operating System: Unknown/Multiple | Architecture: | Unknown/Multiple Type of failure: None/Unknown | Test Case: Blocked By: | Blocking: Related Tickets: | Differential Rev(s): Phab:D1349 Wiki Page: | -------------------------------------+------------------------------------- Changes (by rwbarton): * status: closed => new * resolution: fixed => Comment: I didn't understand Duncan's "ulterior motives" before, but since he just mentioned them on IRC, I'm rather sure they are wrong. We should create a `MIN_VERSION_*` macro for every ''visible'' package. Cabal builds with `-hide-all-packages` anyways, so this is no behavior change compared to what's currently implemented when building with Cabal. The performance worry seems suspect, too. Besides the fact that Cabal builds are still unaffected, the amount of work needed to write out the header file can't be large in comparison to the time needed to read the package database in the first place. The only non-linear case would be using `ghc --make` to build a large number of modules while having a large number of packages in the package database, when cpp would have to process the header file once for every package. I think this is still unlikely to be a big issue compared to ghc actually compiling the modules. I'll try to do some timing tests. On general principle, it seems bad to unnecessarily create multiple "tiers" of visible packages: the visible packages but also the ''really'' visible packages, that were specified explicitly. -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/10970#comment:7 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler

#10970: Built in MIN_VERSION macro support -------------------------------------+------------------------------------- Reporter: ezyang | Owner: Type: feature request | Status: new Priority: normal | Milestone: 8.0.1 Component: Compiler | Version: 7.11 Resolution: | Keywords: Operating System: Unknown/Multiple | Architecture: | Unknown/Multiple Type of failure: None/Unknown | Test Case: Blocked By: | Blocking: Related Tickets: | Differential Rev(s): Phab:D1349 Wiki Page: | -------------------------------------+------------------------------------- Comment (by ezyang): Well, it's literally a two line change. I'm willing to flip the switch. Anyone else? -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/10970#comment:9 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler

#10970: Built in MIN_VERSION macro support -------------------------------------+------------------------------------- Reporter: ezyang | Owner: Type: feature request | Status: new Priority: normal | Milestone: 8.0.1 Component: Compiler | Version: 7.11 Resolution: | Keywords: Operating System: Unknown/Multiple | Architecture: | Unknown/Multiple Type of failure: None/Unknown | Test Case: Blocked By: | Blocking: Related Tickets: | Differential Rev(s): Phab:D1349, Wiki Page: | Phab:D1869 -------------------------------------+------------------------------------- Changes (by thomie): * differential: Phab:D1349 => Phab:D1349, Phab:D1869 -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/10970#comment:11 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler

#10970: Built in MIN_VERSION macro support -------------------------------------+------------------------------------- Reporter: ezyang | Owner: Type: feature request | Status: new Priority: normal | Milestone: 8.0.1 Component: Compiler | Version: 7.11 Resolution: | Keywords: Operating System: Unknown/Multiple | Architecture: | Unknown/Multiple Type of failure: None/Unknown | Test Case: Blocked By: | Blocking: Related Tickets: | Differential Rev(s): Phab:D1349, Wiki Page: | Phab:D1869 -------------------------------------+------------------------------------- Comment (by rwbarton): Replying to [comment:12 hvr]:

Replying to [comment:3 rwbarton]:

...

[...] I don't understand any more why it should be the case that we only create these macros for packages explicitly specified with `-package` or equivalent.

Well, if you `-hide-all-packages` and unmask a few packages via `-package`, then exposing macros for all other packages in the package database would be very unsound, as it would destroy modularity.

I was unclear in comment:3. I thought I was pretty clear in comment:7, but let me try again. For a given compilation, after processing all package-related flags and environment variables, and reading in all package databases, GHC selects some set of installed packages, called the exposed packages. These are the packages that GHC will use to satisfy import statements. My proposal is simply: define `MIN_VERSION_x` macros for exactly the exposed packages. If you can import it, then you can check its version too. So, if you `-hide-all-packages` and unmask a few packages via `-package`, you'd only get `MIN_VERSION_x` macros for those packages you exposed with `-package`. The modularity concerns that Cabal addresses for imports then extend directly to `MIN_VERSION_x` macros. In Cabal, you need to state your dependency on a module to import it. So, you need to state your dependency on a module to check its version with `MIN_VERSION_x`, too. No need to reason explicitly about `-hide-all-packages` and `-package` flags to arrive at this result. -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/10970#comment:13 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler

#10970: Built in MIN_VERSION macro support -------------------------------------+------------------------------------- Reporter: ezyang | Owner: Type: feature request | Status: patch Priority: normal | Milestone: 8.0.1 Component: Compiler | Version: 7.11 Resolution: | Keywords: Operating System: Unknown/Multiple | Architecture: | Unknown/Multiple Type of failure: None/Unknown | Test Case: Blocked By: | Blocking: Related Tickets: | Differential Rev(s): Phab:D1349, Wiki Page: | Phab:D1869 -------------------------------------+------------------------------------- Comment (by hvr): Replying to [comment:14 bgamari]:

This (Phab:D1869) will be merged when #11763 is implemented.

#11763 has been implemetned already... but we should make sure that the Cabal 1.24 bundled with GHC 8 is updated to use `-fno-version-macros` so that Cabal controls 100% of which macros are made available to programs; it's a bad idea IMHO to have both, `cabal_macros.h` and `-fversion-macros` in effect even if it doesn't cause apparent problems *yet*. -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/10970#comment:15 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler

#10970: Built in MIN_VERSION macro support -------------------------------------+------------------------------------- Reporter: ezyang | Owner: Type: feature request | Status: patch Priority: normal | Milestone: 8.0.1 Component: Compiler | Version: 7.11 Resolution: | Keywords: Operating System: Unknown/Multiple | Architecture: | Unknown/Multiple Type of failure: None/Unknown | Test Case: Blocked By: | Blocking: Related Tickets: | Differential Rev(s): Phab:D1349, Wiki Page: | Phab:D1869 -------------------------------------+------------------------------------- Comment (by Ben Gamari ): In [changeset:"e9c2555ac666912f7dff56448ced4bfa06d14e76/ghc" e9c2555/ghc]: {{{ #!CommitTicketReference repository="ghc" revision="e9c2555ac666912f7dff56448ced4bfa06d14e76" Don't require -hide-all-packages for MIN_VERSION_* macros Define MIN_VERSION_pkgname and VERSION_pkgname macros for all exposed packages, without requiring -hide-all-packages. See #10970 comment 7-10 for discussion. Reviewers: duncan, ezyang, bgamari, austin Reviewed By: ezyang Subscribers: hvr, rwbarton Differential Revision: https://phabricator.haskell.org/D1869 GHC Trac Issues: #10970 }}} -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/10970#comment:17 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler