clarifying Availability: Unix

[sandrotosi]

BPO	11233
Nosy	@freddrake, @birkenfeld, @pitrou, @vstinner, @ezio-melotti, @merwok, @bitdancer, @sandrotosi, @berkerpeksag, @csabella
PRs	bpo-11233: Create availability directive for documentation #9692 [3.7] bpo-11233: Create availability directive for documentation (GH-… #9830
Files	availability-directive.patch

^{Note: these values reflect the state of the issue at the time it was migrated and might not reflect the current state.}

Show more details

GitHub fields:

assignee = None
closed_at = <Date 2018-10-15.21:14:27.927>
created_at = <Date 2011-02-17.19:03:18.252>
labels = ['3.8', 'type-feature', '3.7', 'docs']
title = 'clarifying Availability: Unix'
updated_at = <Date 2018-10-15.23:02:03.269>
user = 'https://github.com/sandrotosi'

bugs.python.org fields:

activity = <Date 2018-10-15.23:02:03.269>
actor = 'cheryl.sabella'
assignee = 'docs@python'
closed = True
closed_date = <Date 2018-10-15.21:14:27.927>
closer = 'vstinner'
components = ['Documentation']
creation = <Date 2011-02-17.19:03:18.252>
creator = 'sandro.tosi'
dependencies = []
files = ['32065']
hgrepos = []
issue_num = 11233
keywords = ['patch']
message_count = 19.0
messages = ['128746', '128750', '128751', '128752', '128833', '128854', '128864', '139140', '141970', '199610', '199623', '263744', '327017', '327020', '327025', '327591', '327784', '327787', '327801']
nosy_count = 12.0
nosy_names = ['fdrake', 'georg.brandl', 'pitrou', 'vstinner', 'ezio.melotti', 'eric.araujo', 'r.david.murray', 'sandro.tosi', 'docs@python', 'python-dev', 'berker.peksag', 'cheryl.sabella']
pr_nums = ['9692', '9830']
priority = 'normal'
resolution = 'fixed'
stage = 'resolved'
status = 'closed'
superseder = None
type = 'enhancement'
url = 'https://bugs.python.org/issue11233'
versions = ['Python 3.7', 'Python 3.8']

[sandrotosi]

Hi, following up http://mail.python.org/pipermail/docs/2011-February/003083.html we have a chat on #python-dev on the topic, the situation is

• all started with os.lchmod() and as.lchflags() methods not available on a Debian system
• that's because POSIX does not require those functions and then Linux systems don't provide them
• so we then thought about clarifying, only for those 2 methods, that they might not be available with a format like "availability: some unix systems"
• then we noticed that on top of os doc there's a notice "An “Availability: Unix” note means that this function is commonly found on Unix systems. It does not make any claims about its existence on a specific operating system." and so we thought about linking every 'Availability: Unix' to that note
• or transform that not in a footnote and link every 'Availability: Unix' for functions that might not be present to that footnote.

It's enough for the conversation dump: ideas/suggestions? :)

[bitdancer]

I like the footnote idea.

I wonder if it would also be worth marking which functions are Posix (Availability: posix) and therefore likely to be available on all unix systems.

[pitrou]

I like the footnote idea.

Me too.

I wonder if it would also be worth marking which functions are Posix
(Availability: posix) and therefore likely to be available on all unix
systems.

POSIX has optional APIs, hasn't it?

[bitdancer]

Good point :(

[birkenfeld]

The "Notes on availability" bullet list at the top of docs.python.org/library/os should already say everything that there is to say here...

[bitdancer]

Yes, but when you jump directly to one of the functions, you don't see that bullet list. Whereas if it were a footnote on the 'avaiability: unix" statement, the natural thing would be to follow the footnote and thus learn the limitations on that statement. Thus I think the footnote would be better than the bullet list at the top.

[birkenfeld]

That's a good point, indeed.

[bitdancer]

In another issue Georg came up with the idea of creating an availability directive that would auto link to the appropriate text. This would then apply to all our different availability types.

[merwok]

If no-one else has started on this, I’m interested in making a patch to introduce an availability directive.

[birkenfeld]

A draft is attached. The directive currently just always links to the "availability" notes, which I have moved to the main "Operating system services" document. Please review!

[pitrou]

I can't comment on the Sphinx extension code but this is a good idea.

[vstinner]

I like the idea of a single place where the Availability is described, and adding a link to this place. availability-directive.patch LGTM.

[csabella]

Hello,

It doesn't appear that this patch was ever merged. If there's still interest, would it be OK for me to convert it to a PR?

Thanks!

[freddrake]

A PR for this would be good, and would certainly accelerate getting this accomplished. Thanks, Cheryl!

[csabella]

Thanks, Fred. I've submitted the PR.

[vstinner]

New changeset 2d6097d by Victor Stinner (Cheryl Sabella) in branch 'master':
bpo-11233: Create availability directive for documentation (GH-9692)
2d6097d

[vstinner]

New changeset b248a8c by Victor Stinner (Cheryl Sabella) in branch '3.7':
[3.7] bpo-11233: Create availability directive for documentation (GH-9692) (GH-9830)
b248a8c

[vstinner]

Thank you Sandro Tosi for the bug report, thanks Georg Brandl for initial patch, and thanks Cheryl Sabella for the final changes in 3.7 and master!

According to Yury Selivanov, it's not need to backport this change to 2.7 and 3.6, so I close the issue.
#9692 (comment)

[csabella]

And thank you, Victor, for reviewing and merging! :-)

[calestyo]

Just wondered... was that really ever fixed? Current documentation still seems to contain no indication that e.g. lchmod is not available on linux.

[merwok]

Not sure what you’re asking. We can see in the timeline on this page that a change was merged, which makes docs such as https://docs.python.org/3/library/os.html#os.lchmod contain a link to https://docs.python.org/3/library/intro.html#availability that clarifies what «availability» means – the goal of this ticket!

[calestyo]

Well, yes... but that's a pretty abstract information, isn't it? It's even a bit unusual, the first point says "Unix doesn't claim that a function exists on a specific OS"... while the second point says "Unix means it is available on MacOS, because that's a Unix".

Anyway... many functions do seem to have specific notes about Linux availability (e.g. when a minimum kernel version is required).

Wouldn't it make sense to notice if a function is specifically not available on Linux?

[merwok]

How could that be said? Doesn’t each feature depend on specifc kernel version and build options?

Edit: I guess there are many that are introduced by freebsd or aix or macos and don’t exist at all on linux. Do you have a list of such functions? If so, please open a new ticket so that this can be corrected!

[calestyo]

I'd have naively guessed that lchmod is simply generally not available in Linux... probably because Linux (and IIRC most other Unixes) ignores any file permission bits on symlinks (thus lchmod doesn't really make sense for it).

It's also not documented in the "Linux manpages"... so I'd say it's a general thing.

[calestyo]

Just found https://unix.stackexchange.com/questions/224979/why-do-linux-posix-have-lchown-but-not-lchmod ... so seems to be as I've suspected before.

[calestyo]

Opened a new issue #107959 and filed an MR.