Re: [PATCH v2] environ.7: align PWD with the standard

[Alejandro Colomar <alx.manpages@gmail.com>]

Hi Stefan, and наб,

On 6/21/22 23:35, Stefan Puiu wrote:
> On Mon, Jun 20, 2022 at 6:23 PM наб <nabijaczleweli@nabijaczleweli.xyz> wrote:
>>
>> Hi!
>>
>> On Mon, Jun 20, 2022 at 11:55:18AM +0300, Stefan Puiu wrote:
>>> On Mon, Jun 20, 2022 at 2:40 AM наб <nabijaczleweli@nabijaczleweli.xyz> wrote:
>>>>
>>>> Signed-off-by: Ahelenia Ziemiańska <nabijaczleweli@nabijaczleweli.xyz>
>>>> ---
>>>>   man7/environ.7 | 8 ++++++--
>>>>   1 file changed, 6 insertions(+), 2 deletions(-)
>>>>
>>>> diff --git a/man7/environ.7 b/man7/environ.7
>>>> index 019c5a25a..24446c709 100644
>>>> --- a/man7/environ.7
>>>> +++ b/man7/environ.7
>>>> @@ -158,8 +158,12 @@ used by
>>>>   to find manual pages, and so on.
>>>>   .TP
>>>>   .B PWD
>>>> -The current working directory.
>>>> -Set by some shells.
>>>> +Absolute path to the current working directory;
>>>> +required to be partially canonical (no
>>>> +.I .\&
>>>> +or
>>>> +.I ..\&
>>>> +components).
>>>
>>> If any shell decides to ignore that part of the spec, is there
>>> anything preventing them?
>> It no longer being a valid shell (if on startup) or providing an invalid
>> cd built-in (when cding), from the stand-point of conformance to both
>> the standard and historical shells.
> 
> My expectation from the Linux manual pages is that they document
> behavior I'm likely to encounter in the real world on Linux, with
> various libcs etc. Specs can be misread, misunderstood, ignored, or
> can be wrong sometimes (see the discussion about fork being required
> to be async signal safe, for example, the RH page here:
> https://access.redhat.com/articles/2921161). So if I write software
> that, say, does getent("PWD"), it's useful to know if there are cases
> where that might not work. Even if POSIX requires PWD to be set,
> that's not reassuring when my program crashes. That's why I asked if
> you checked some shells to see what they do.
> 
> Also, I see there's already environ.3p for the POSIX version.
> 
>>
>>> I would make it clear in the text that this
>>> is a spec requirement, not a practical guarantee (e.g. "required by
>>> <spec> to be ...").
>> Those are one and the same, that's the point of SUS/POSIX
>> (and conformance to historical implementations).
>> Are you aware of one or are you just concern-trolling?
> 
> I don't know of a shell that doesn't set PWD, but since you said the
> previous comment ("Some shells set it") was wrong, I assumed you had
> checked that. I did try bash and busybox sh and they seem to set it,
> but there are (quite) a few other shells out there.


Since no-one seems to know of a shell that deviates from this behavior, 
and no-one replied in a couple of weeks, I'm going to assume that all 
well-known shells follow POSIX, and that the wording was falling on the 
safe side of things in the old times where following standards was not a 
common thing to do (or maybe the page was written before there were any 
standards).

So I applied the patch.

> 
> I don't know why you think that my question is some kind of trolling.
> You can look up my contributions in the mailing list archive. It's
> been a while since my last patch to the project; nowadays I follow
> some of the mailing list discussions in my spare time and occasionally
> chime in on people's patches. I get things wrong sometimes, of course,
> but your reply is the first one that is defensive (and somewhat rude,
> I would say). Probably a good indication that I need to find other
> uses of my (spare) time.

Thanks for following the list, I appreciate your reviews!

> 
>> Obviously, pretty much no part of this manual applies to csh
>> because csh is its own 2BSD brand of insanity (in this case
>> largely because it predates V7 (3BSD), and, hence, the environment).
>> csh users understand they use a non-shell,
> 
> Well, this is the environ.7 man page, mostly useful to programmers
> AFAICT, who don't have much control over what people use as their
> shell. If they can set csh as their login shell, why should I assume
> they won't? People have many reasons to choose the shell they run -
> distro default, they like some features, speed, company policy, legacy
> systems... Not sure POSIX compliance is high on that list.

I checked csh(1) provides PWD, at least on my system (and in fact, it's 
the only variable it sets):

$ env -i csh
% env
PWD=/home/alx
% csh -l
% env
PWD=/home/alx
%

I'm going to assume that the text is not true, unless someone reports 
otherwise.

Thanks you both!

Alex
-- 
Alejandro Colomar
<http://www.alejandro-colomar.es/>