Ticket #619 (closed roadmap: wontfix)

Opened 5 years ago

Last modified 3 years ago

eliminate perl 5 dependency

Reported by: allison Owned by:
Priority: trivial Milestone: 3.0
Component: configure Version:
Severity: medium Keywords:
Cc: Language:
Patch status: Platform:

Description

Eliminate Parrot's configuration and build dependency on Perl 5. This roadmap task has several components (which might be split into separate tickets later), including:

- A PMC preprocessor written in PGE/PCT instead of Perl 5 regular expressions. (The generated C files can be checked into the repository and shipped in the tarball.)

- A configuration system that doesn't use data from the system install of Perl, and doesn't use Perl during configuration and build. It's okay to require Perl on the developer's system, like Autoconf. (Bonus for using an existing configuration system like Autoconf rather than rewriting one from scratch.) See ParrotBasedConfig.

- A testing infrastructure written in Parrot (in PIR, or an HLL compiled down to PIR), so testing Parrot doesn't depend on Perl 5. The new testing infrastructure will likely continue to use TAP, though other formats for result reporting will be considered.

Change History

  Changed 5 years ago by jkeenan

For a contrarian view, see  this post by David Golden.

  Changed 5 years ago by jkeenan

  • component changed from none to configure

  Changed 4 years ago by coke

  • priority changed from normal to trivial

follow-up: ↓ 5   Changed 4 years ago by dukeleto

I see value in getting rid of unnecessary dependencies, and I see the value porting some tests that are currently in Perl 5 to PIR, but converting our entire configure, build and test subsystems is a huge undertaking. What is the real gain?

I totally understand there are many bugs that are related to how we figure out compile time flags from Perl's build flags, but that is orthogonal to totally removing Perl 5 from the above subsystems.

I think this ticket needs to be broken down into smaller, doable chunks, with the benefit of each described.

in reply to: ↑ 4   Changed 4 years ago by jkeenan

Replying to dukeleto:

I see value in getting rid of unnecessary dependencies, and I see the value porting some tests that are currently in Perl 5 to PIR, but converting our entire configure, build and test subsystems is a huge undertaking. What is the real gain? I totally understand there are many bugs that are related to how we figure out compile time flags from Perl's build flags, but that is orthogonal to totally removing Perl 5 from the above subsystems.

Along the same lines, I'd like to quote a comment made by Andy Dougherty three months back in TT #1044:

I have never understood TT #619, particularly the bit about replacing 
reasonably portable Perl with a Unix shell script (and, presumably, 
parallel sets of hand-maintained config files for any non-unix platforms, 
such as Windows).

  Changed 3 years ago by dukeleto

  • status changed from new to closed
  • resolution set to wontfix

Consensus is -1 to do this. If a subset of this ticket is wanted, please create a Github issue:

 https://github.com/parrot/parrot/issues/new

Note: See TracTickets for help on using tickets.