Re: 01/01: services: Add ‘/usr/bin/env’ special file.

[Jesse Gibbons]

Hi T-G-R,
On Sat, 2019-09-07 at 09:52 +0200, Tobias Geerinckx-Rice wrote:
> Jesse,
> 
> Thanks!  It was linked from another thread[0] Ludo' pasted to my 
> patch, though.  I've read both.
> 
> Jesse Gibbons 写道：
> > Here's a post with what I think is a good argument against 
> > adding
> > /usr/bin/env. I think the standard patch-shebang phase does a 
> > good job
> > at preventing the potential issue, but the argument still 
> > applies to
> 
> So is the argument here that Guix packages would generate scripts 
> at run-time that refer to /usr/bin/env?  Or just that it's 
> currently easy to catch packages that install #!/usr/bin/env 
> scripts?  Or something else?
> 
> Neither /bin/sh nor /usr/bin/env are available in the build 
> environment.  Relying on the *likely* run-time non-existence of 
> /usr/bin/env in *user* environments to catch *packaging* bugs does 
> not sound acceptable to me.
> 
> > scripts generated by make, as I learned while attempting to port
> > pysolfc.
> 
> If you have the time, could you elaborate?
It's a bit complicated.
Pysolfc has a horrible build system. It doesn't build the required i18n
unless you call "make test". Howevver, "make test" runs a script
generates python test scripts that use #!/usr/bin/env to find both
python and python2. Since they are generated by a shell script, patch-
shebangs does not catch them, and it fails to find /usr/bin/env. It's a
mess that I don't know how to solve.

I've attached my pysolfc.scm so you can observe what I mean. Sorry it's
a horrible mess of commented broken code.
> 
> Kind regards,
> 
> T G-R
> 
> [0]: https://debbugs.gnu.org/cgi/bugreport.cgi?bug=35910

-- 
-Jesse

pysolfc.scm
Description: Text Data