I don't think anyone has proposed a tool like virtualenv, so I don't see how that's relevant. And I don't see why we should make impositions on the way anyone chooses to package j.

On Fri, 10 Mar 2023, Raul Miller wrote:

Yes, PATH is the only environmental variable that influences how
/usr/bin/env finds the executable.

But I was talking about tools like
https://pypi.org/project/virtualenv/ where there's other things going
on.

That said, you are correct that there are some systems which install
optional software to /usr/local/bin/, and I had forgotten about those.
That said... I don't think there's any significant permissions
problems involved in installing in /usr/bin on bsd systems. As I
understand it, that's /usr/bin/env bash is a historical issue (and,
thus, a backwards compatibility issue).

Thanks,

--
Raul

On Fri, Mar 10, 2023 at 8:20 PM Elijah Stone <elro...@elronnd.net> wrote:

> /usr/bin/env is the standard for developer systems where different
> configurations are managed through environmental variables. However, it's
> rather fragile for non-developers, as it requires properly conditioned
> environmental variables.

This statement confuses me greatly.  Can you elaborate?  I think the only
environment variable which is relevant is PATH, and PATH will certainly
contain /usr/bin.  So if PATH contains /usr/bin, and there is a
/usr/bin/ijconsole, then #!/usr/bin/env ijconsole will do the right thing.  If
ijconsole is somewhere other than /usr/bin, and it is in PATH, then
#!/usr/bin/env ijconsole will pick it up, but #!/usr/bin/ijconsole will not.
So using /usr/bin/env is strictly less fragile.

Using /usr/bin/env is standard for scripts.  /usr/bin/ijconsole entails
_strictly_ more fuss, more fragility, and less portability.

An example: on many BSD systems, bash does not come with the base system, but
is rather provided as a separate package, so it is placed in /usr/local/bin.
Some scripts developed on linux are distributed that use #!/bin/bash, and this
breaks on BSD systems because there is on bash in /bin; when those scripts use
#!/usr/bin/env bash instead, they work.

On Fri, 10 Mar 2023, Raul Miller wrote:

>
> /usr/bin/$name is the earlier standard for minimal fuss distribution.
>
> Given that most J users are going to be in some IDE, I don't see that
> the /usr/bin/env approach would provide significant benefit to the J
> community. Especially if we're considering people who would be baffled
> about what to do with #!/usr/bin/ijconsole
>
> Do you think I've overlooked something important here?
>
> Thanks,
>
> --
> Raul
>
> On Fri, Mar 10, 2023 at 7:42 PM Elijah Stone <elro...@elronnd.net> wrote:
>>
>> Performance is irrelevant here--the difference in either case is miniscule.
>>
>> Using /usr/bin/env is standard practice in shebangs for finding binaries on
>> the path.  In some systems, it is conventional to put binaries that did not
>> come with the system in /usr/local/bin, or in /opt; whereas, the existence of
>> /usr/bin/env is mandated by posix.  We should not presume any more than we
>> need to about the way a user's system will be organised.
>>
>> On Fri, 10 Mar 2023, Raul Miller wrote:
>>
>> > Yes, the name ijconsole (iverson j console) was adopted to work around
>> > java's jconsole being in $PATH on many systems.
>> >
>> > This would also be an issue with /usr/bin/env
>> >
>> > But I don't see any high priority problems with /usr/bin/ijconsole as
>> > a standard location. There will be some users who cannot install J
>> > there, but it should be simple enough for them to build a tool to
>> > patch incoming J scripts with the install location that they used.
>> >
>> > (And, /usr/bin/env is going to be considerably slower than a bit of
>> > undefined verb parsing and garbage collection within J.)
>> >
>> > --
>> > Raul
>> >
>> > On Fri, Mar 10, 2023 at 6:51 PM Elijah Stone <elro...@elronnd.net> wrote:
>> >>
>> >> Fair enough.  I think the obvious thing is #!/usr/bin/env jconsole.  But I
>> >> gather 'jconsole' is also the name of a java-related program, which might 
be
>> >> why debian uses 'ijconsole' (and I think I have seen 'jc' 
elsewhere--probably
>> >> for the same reason).  I therefore propose:
>> >>
>> >> 1. Rename jconsole to something which doesn't collide with anything 
heretofore
>> >> notable
>> >>
>> >> 2. Use #!/usr/bin/env that-something
>> >>
>> >> On Fri, 10 Mar 2023, Raul Miller wrote:
>> >>
>> >> > Perhaps, but I think it would be more useful to have a portable,
>> >> > agreed on path for systems which support the #! convention.
>> >> >
>> >> > That's required for J scripts to be distributable.
>> >> >
>> >> > Thanks,
>> >> >
>> >> > --
>> >> > Raul
>> >> >
>> >> > On Fri, Mar 10, 2023 at 12:04 AM Elijah Stone <elro...@elronnd.net> 
wrote:
>> >> >>
>> >> >> Perhaps it's simply time to tell the interpreter to start ignoring the 
first
>> >> >> line of a script if it starts with #!...
>> >> >>
>> >> >> On Thu, 9 Mar 2023, Raul Miller wrote:
>> >> >>
>> >> >> > The recent change in directory naming from j903 to j9.4 introduces an
>> >> >> > interesting issue for shell scripts on unix-like systems.
>> >> >> >
>> >> >> > In J shell scripts, this works:
>> >> >> >
>> >> >> > #!/home/username/j903/bin/jconsole
>> >> >> >
>> >> >> > However, this fails with a spelling error:
>> >> >> >
>> >> >> > #!/home/username/j9.4/bin/jconsole
>> >> >> >
>> >> >> > Or, on OSX, the shebang line is different, but the spelling error 
remains:
>> >> >> >
>> >> >> > #!/Applications/j9.4/bin/jconsole
>> >> >> >
>> >> >> > That said, there's some other issues here, related to portability. 
One
>> >> >> > of which is that (as a general rule) a home directory is personal
>> >> >> > rather than portable. Another is that there's official java jconsole
>> >> >> > which does not understand J.
>> >> >> >
>> >> >> > But, also, for a long time now, Debian based distributions have been
>> >> >> > distributing j with /usr/bin/ijconsole as a symbolic link to the
>> >> >> > current installed location for J. (Here, the 'i' in ijconsole stands
>> >> >> > for Iverson.)
>> >> >> >
>> >> >> > So I think that now would  be a good time to adopt that as 
"documented
>> >> >> > standard practice" for j shell scripts.
>> >> >> >
>> >> >> > FYI,
>> >> >> >
>> >> >> > --
>> >> >> > Raul
>> >> >> > 
----------------------------------------------------------------------
>> >> >> > For information about J forums see 
http://www.jsoftware.com/forums.htm
>> >> >> ----------------------------------------------------------------------
>> >> >> For information about J forums see http://www.jsoftware.com/forums.htm
>> >> > ----------------------------------------------------------------------
>> >> > For information about J forums see http://www.jsoftware.com/forums.htm
>> >> ----------------------------------------------------------------------
>> >> For information about J forums see http://www.jsoftware.com/forums.htm
>> > ----------------------------------------------------------------------
>> > For information about J forums see http://www.jsoftware.com/forums.htm
>> ----------------------------------------------------------------------
>> For information about J forums see http://www.jsoftware.com/forums.htm
> ----------------------------------------------------------------------
> For information about J forums see http://www.jsoftware.com/forums.htm
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm

Reply via email to