Re: [gpsd-dev] Planning a 3.11 snap release

[Greg Troxel]

address@hidden (Eric S. Raymond) writes:

> * We need to make a final design-level decision about the leapfetch code.
>   Presently it's enabled by default, but it has a clash with 
>   parallel builds. Keep it this way?  Default to leapfetch=no?
>   Remove it entirely?  

I think fetching things at build time is basically broken.  Two reasons
argue in favor of this, besides it being tradtional:

  packaging systems have a fetch stage and build stage.  The build stage
  should not use the net; part of the point of separation is to be able
  to run fetch in one place and transport the distfiles.

  Builds should be repeatable.  With the same toolchain and
  dependencies (and hacks to set timestamps to 0), one should get the
  same binary package from the same source tarball.

I think leapfetch should be removed from the *build* process entirely.
It's certainly reasonable to have a "secondary fetch" step to freshen
it, that can be run to commit a new file into git, or by someone who
chooses to do that.  If there are releases every 6 months, then this
isn't that big a deal anyway.

pgpJ1xiPOgOhR.pgp
Description: PGP signature