Attn: Install path debaters

Gary E. Miller gem at rellim.com
Wed Jan 3 22:44:48 UTC 2018


Yo Ian!

On Wed, 3 Jan 2018 16:25:34 -0600
Ian Bruene via devel <devel at ntpsec.org> wrote:

> > Uh, news to me that any solution was agreed to.  Last I heard this
> > group was in no way on the same page.
> >
> > Rather than having me misread your code, can you put a plain
> > summary here?  
> 
> It's rlaager's code, the bash sys.path in each program one.

Then maybe rlaager can provide a summary of the 4 different cases:

	A. distro install: /etc, /bin, etc.
        B. system install: /usr/local/etc, /usr/local/etc, etc.
        C. user install: ~/etc, ~/etc, etc. or similar
        D. packager install: /tmp/etc, /tmp/bin, going into a package

And, most important, in some systems, all FOUR must be able to
coexist.
        
> But since September everyone has been locked in a loop:

Yup.  Because we really need 4 totally different solutions, people
keep trying to get to 1.

No otehr projects seem to have this loop issue...

> 1. Violate FHS. If this is kosher then it is also kosher to mangle
> the user's PYTHONPATH and we should do that instead.

Only for B. and C.  No other packages seem to need to violate the FHS.
I agree PYTHONPATH is the standard solution used by all.

> 2. Old system of modules randomly inaccessible. (I would be tolerant
> of this)

Uh, lost me.  What old system?  NTPsec has been using PYTHONPATH just
fine.

> 3. .pth files. I'm not clear about why these are horribly broken, but 
> they must be or they would have not been shot down as soon as they
> were mentioned.

Once again, only usefull for SOME cases of B. and C.  Violates the
FHS.  Breaks many use cases where A., B., C., and/or D. co-exist.


> 4. The nightmare of !615

I've not looked at the code.  Care to elaborate?

> 5. Have I forgotten something?

Taht is a given.  :-)

> > And yet, other projects do not have this problem???  
> 
> This is not helpful. Someone needs to give a concrete example of
> another project that achieves this (and where the solution is
> compatible with waf) so we can see what they are doing.

Do not conflate two different issues:
	1. where and how to install
        2. host to get waf to do 1.

I have taken no position on #2 as we have not agree on #1.

No point thinking about how to do #2 until #1 is agreed.

Is it more helpfuil if I mention that gpsd has no issues with A., B., C.,
and D.  gpsd does the totally standard python install thing, just like
ntpsec needs to.

> > Care to shaare that ccomment so I do not need to dig???  
> 
> Not yet written. It would summarize the issues we have had and why we 
> picked the particular solution.

Well then, that needs to be done.  If we can't agree on the summary then
the 615 code can't be merged.

RGDS
GARY
---------------------------------------------------------------------------
Gary E. Miller Rellim 109 NW Wilmington Ave., Suite E, Bend, OR 97703
	gem at rellim.com  Tel:+1 541 382 8588

	    Veritas liberabit vos. -- Quid est veritas?
    "If you can’t measure it, you can’t improve it." - Lord Kelvin
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 488 bytes
Desc: OpenPGP digital signature
URL: <https://lists.ntpsec.org/pipermail/devel/attachments/20180103/bfa6c0c6/attachment.bin>


More information about the devel mailing list