bpo-41818: Make additions to the tty module by 8vasu · Pull Request #23546 · python/cpython · GitHub

This follows #23536. Also, see #23686, #23740.

Add cfmakeraw(), cfmakeecho(), and cfmakecbreak(); make setraw() and setcbreak() return saved termios attributes to avoid extra tcgetattr() calls.

Post #23686, #23546, #23740 goals:

add test_winsize() to "Lib/test/test_pty.py";
major revision of "Lib/pty.py", which has not happened since the beginning of the millennium.

Notes

List of all POSIX.1-2017 input mode flags - https://pubs.opengroup.org/onlinepubs/9699919799/basedefs/termios.h.html#tag_13_72_03_02.
List of all POSIX.1-2017 local mode flags - https://pubs.opengroup.org/onlinepubs/9699919799/basedefs/termios.h.html#tag_13_72_03_06.
Naming justification for the functions: https://man7.org/linux/man-pages/man3/cfmakeraw.3.html Note that cfmakeraw() is not POSIX compliant; it is available on the BSDs and in glibc. However, POSIX has functions with similar names; for example, consider cfgetispeed(). Please see https://pubs.opengroup.org/onlinepubs/9699919799/basedefs/termios.h.html.
test_winch() will be resubmitted as a part of a later pull request.
bpo-41818: Add termios.tcgetwinsize(), termios.tcsetwinsize(). Update docs. #23686 replaces class winsize.
bpo-41818: Add os.login_tty() for *nix. #23740 replaces tty.login_tty().

Signed-off-by: Soumendra Ganguly soumendraganguly@gmail.com

https://bugs.python.org/issue41818

@asvetlov @ethanfurman this follows #23536.

TODO: update documentation for the tty module; maybe also add tests for it.

Note: I have put my name in one of the files; let me know if that is not customary :D

I suggest esablish public function names first, update the tty module documentation, and make the final review.

@jonathanslenders I very respect your work on https://github.com/prompt-toolkit/python-prompt-toolkit
Maybe you want to review this (and follow-up) PR? Maybe you can suggest something? I really like to hear voices from experts in PTY/TTY area.

Duplicate

As you suggested, I will use hasattr() for this instead.

Explicit check for hasattr(termios, "TIOCSWINSZ") or something like this is better than except NameError.

It did feel like there had to be a better way of doing this. I did not like the except NameError either. Thank you so much for this suggestion.

I have one concern. To be able to use hasattr(termios, "TIOCSWINSZ"), I think I need to import termios. Am I correct about this? Since we already have from termios import *, which cannot be removed because of backward compatibility, is it good practice to also have an import termios there just for that single purpose? I was thinking that we could use

try:
   from termios import TIOCGWINSZ, TIOCSWINSZ
   ...
except ImportError:
   ...

which seems to be an explicit check as well.

I think better to make a stub that raises NotImplementedError is better than setting the function name to None.
In std lib we never use this pattern.

Sometimes we don't declare unsupported function at all, but in this case we usually mark in docs: "Availability: Windows" or "Availability: Linux kernel >= 4.5 or glibc >= 2.27".

I think here is better to support stubs that always fail.

I am working on this.

mode_echo is not the best name I guess.
Looking on the function body I can suggest renaming with enable_echo().

enable_echo(mode) and enable_echo(mode, False).
Maybe others have better idea?

enable_echo() does sound better than mode_echo(). If you agree with makeraw() instead of mode_raw(), then we could also change this to makeecho(); however, I think that enable_echo() sounds better than either. I am leaning toward enable_raw(), and the updated commit will use that name unless someone else suggests something better.

How about mkecho()? See next section.

Sounds good

enable_raw() maybe if we agree with enable_echo()?

Yes, that was a temporary name. This is supposed to function like cfmakeraw(3). While enable_raw() does sound better than mode_raw(), we could also change it to tty.makeraw() if you permit me to do so, and if there are no licensing issues.

How about mkraw() instead of makeraw()? Reasoning: mkecho() seems better than makeecho().

Ok

I have changed the names to cfmakeraw(), cfmakeecho(), ... so that they conform to standard naming convention.

The naming again.
For me, the function does not login but makes the calling process a session leader as you correctly wrote in docstring.

Maybe make_session_leader() or set_session_leader() is better?

Perfect! I will use one of those, or a variant of one of those. I agree that tty.login() was poor choice on my part; I wanted it to sound like login_tty(3).

Update: since I had written these a while ago, I had forgotten what this function exactly does. I did some research, and it turns out that this function is actually doing a bit more than making the calling process a session leader.

According to https://pubs.opengroup.org/onlinepubs/9699919799/functions/setsid.html, the setsid() call in our function is responsible for making the calling process a session leader; moreover

Upon return the calling process shall be the session leader of this new session, shall be the process group leader of a new process group, and shall have no controlling terminal.

Our function is also making fd the stdin/stdout/stderr and controlling tty of the calling process, which is why it belongs in the tty module.

Nevertheless, as you said, login() sounds like a bad choice; I completely agree with that. Should I change it to login_tty()? This function IS actually a clone of login_tty(3).

Additional note: on my Debian system, man login_tty says

These are BSD functions, present in glibc. They are not standardized in POSIX.

Maybe we should add pty.login_tty() function instead which calls the native login_tty(3)?
It can do some platform-specific things which we don't want to learn and reimplement in Python code.

Sir, this sounds like a great idea. I have removed tty.login_tty() from this pull request. I have also done restructuring/cleanup. Please check the comments on the diff.

Please see #23740 for my implementation of os.login_tty().

A Python core developer has requested some changes be made to your pull request before we can consider merging it. If you could please address their requests along with any other requests in other reviews from core developers that would be appreciated.

Once you have made the requested changes, please leave a comment on this pull request containing the phrase I have made the requested changes; please review again. I will then notify any core developers who have left a review that you're ready for them to take another look at this pull request.

@asvetlov Thank you very much for the reviews. As per your suggestion, the updated commit will include documentation, etc.

Edit: you mentioned "I suggest esablish public function names first, update the tty module documentation, and make the final review". I am inexperienced when it comes to the cpython development process. How do I "establish public function names"? Thank you in advance.

@asvetlov : Thanks!

A few thoughts:

For consistency, don't we need an mkcbreak as well?
The setraw implementation in prompt_toolkit is slightly different. But I'm certainly not an expert on this.

See: https://github.com/prompt-toolkit/python-prompt-toolkit/blob/master/prompt_toolkit/input/vt100.py#L204
It's a context manager.
It's clearing IXOFF as well (not sure if needed).
We clear INLCR and IGNCR to avoid translation of carriage return to newlines on input. This patch only clears ICRNL.
There are a few other flags we don't clear in prompt_toolkit, but probably should: BRKINT, INPCK, ISTRIP, OPOST, CSIZE, PARENB.

I'm not familiar with the login_tty functionality.
setwinsize and getwinsize look fine. I've never used the pixel values though, and I've never done a "set size".

Overall I think this looks good.

One thing slightly related, but out of scope for this PR, that often missed in Python is that sys.stdin/sys.stdout are not context variables. So it's impossible to write an asyncssh server that exposes a REPL where print/input is working properly. (I guess I should propose that on python-ideas or bpo, but no energy/time to drive the discussion now.)

@asvetlov Thank you for your patience. I have made the requested changes; please review again. I restructured the files a little bit. As a result, your reviews are showing up as outdated; they are not. Here are some notes to make reviewing easier for you:

Instead of using hasattr(termios, "TIOCGWINSZ"), I used

try:
    from termios import TIOCGWINSZ
    ...
except ImportError:
    ...

I did this to avoid import termios, since we already have a from termios import * which cannot be removed because of backward compatibility reasons. I understand that from termios import * and import termios can coexist; it just felt a little strange doing that :D If you think that using import termios + hasattr() is the better option, then let me know; I will make the changes.

Did I add the stubs for getwinsize() and setwinsize() properly?
I have renamed the functions. One of the comments that I left on one of your reviews was related to the naming of login_tty(); please let me know if that is acceptable.
getwinsize() and setwinsize() now handle the packing/unpacking of structs. As the result, their input/output are tuples of integers instead of bytestrings. That makes them more user-friendly; this helped me remove the pack("HHHH", ...) from test_pty.py and abstract out the important parts.
I have written some documentation. I welcome suggestions!

I too would be honored if this work was reviewed by @jonathanslenders.

Edit: after posting this, I noticed that @jonathanslenders has already posted a review! Thank you very much for this!

Thanks for making the requested changes!

@asvetlov: please review the changes made to this pull request.

Could we return a namedtuple from getwinsize?

I'm not sure what this line is doing exactly? Maybe add a comment here?

Here is the source of login_tty(3): https://sourceware.org/git/?p=glibc.git;a=blob;f=login/login_tty.c

The

    except (NameError, OSError):
        tmp_fd = os.open(os.ttyname(STDOUT_FILENO), os.O_RDWR)

in our code is equivalent to their

 {
        /* This might work.  */
        char *fdname = ttyname (fd);
        int newfd;
        if (fdname)
          {
            if (fd != 0)
              (void) close (0);
            if (fd != 1)
              (void) close (1);
            if (fd != 2)
              (void) close (2);
            newfd = open (fdname, O_RDWR);
            (void) close (newfd);
          }
 }

I contacted the glibc people a while ago and this is what they had to say

Since this was added by Roland almost 25 years ago, it seems to be trying to
emulate the TIOCSCTTY for system derived from System V (as indicated by this
stackflow thread [1]).

Back in the day glibc tries to enclose fallbacks to be portable to the various
Unix implementations. Currently this is just a dead code with various
shortcomings (as for a lot of the glibc fallback implementation where proper
kernel support was the correct answer) since neither Linux nor Hurd uses it.

And it seems that Unixes derived from System V also has moved on from this
interface, neither Solaris nor Illumos seems to provide it.

https://urldefense.com/v3/__https://stackoverflow.com/questions/51593530/code-explanation-for-glibc-login-tty-function-openttyname-immediately-f__;!!KwNVnqRv!Vtu1n8f09GnQ28wafOTRPvyTWxkgO0bd7c3sHMxR5GQqPjq4jJhaL_8J8kw-ZliJ$

TL;DR: TIOCSCTTY is available everywhere now. Therefore, I too think that it would be safe to remove the except (NameError, OSError): block. Should I remove it?

I am going to use tty.login_tty() [ or whatever we end up naming it ] to replace the fallback code in pty.fork(), which still uses the old System V approach. tty.login_tty() will also be used in pty.spawn() to make it easier for us to implement some new/improved functionality.

This comment implies that fd can either be or not be a descriptor of a tty.
Not sure what would happen if it's not a tty descriptor.

You are right. I will have to rephrase that. Actually, it will fail if fd is not a tty. Do you think it would be a good idea to instead mention/document what exception would be raised if fd was not a tty?

@asvetlov @jonathanslenders I have made the requested changes; please review again.

Thanks for making the requested changes!

@asvetlov: please review the changes made to this pull request.

These are some of my old comments and notes; I used them to track progress and make decisions; they may be ignored. Since the comment section was a total mess with mostly me talking to myself, I had to put them together to make it look cleaner. I had also made some "reviews" as notes for myself; they are now gone since I removed a lot of code (the entire class winsize, tty.login_tty() etc) by making a force-push.

old comment 1

@asvetlov @jonathanslenders

Yes! Named tuples sound like a good idea. Working on it!
mkcbreak will be added.
Thank you for pointing us toward the setraw() implementation in prompt_toolkit. I will look into it after taking a nap :D My original idea was to make a clone of cfmakeraw(3): https://sourceware.org/git/?p=glibc.git;a=blob;f=termios/cfmakeraw.c, which does what you mentioned except for clearing INPCK. Should I make it clear all the flags that glibc's cfmakeraw() clears + maybe INPCK as well?

old comment 2

@asvetlov @jonathanslenders I have made the requested changes; please review again. Notes follow.

Updated docs.
mkcbreak() implemented.
Our current mkraw now sets/unsets all flags mentioned by @jonathanslenders except that IXOFF is not being cleared. I think that it would not hurt to clear IXOFF; it is just that cfmakeraw(3) does not do it. Should I go ahead and also clear IXOFF? Since IXOFF is POSIX compliant, it will be present on every system that is immediately relevant.
getwinsize() now returns a namedtuple; the field names are same as that of the traditional struct winsize. The documentation is also advertising the winsz argument of setwinsize() as being a namedtuple of the same type.
A comment has been added to the except (NameError, OSError): clause in tty.login_tty().
Docstring of tty.login_tty() has been modified to reflect the fact that fd must be a file descriptor of a tty.

old comment 3

https://pubs.opengroup.org/onlinepubs/7908799/xcurses/intov.html#tag_001_005_002 mentions that cbreak Mode must additionally have ICRNL cleared. The document is about "xcurses"; however, I checked using stty that even the python curses.cbreak() clears ICRNL. I am going to make another updated commit with very detailed comments!

old comment 4

I compared the following definitions/implementations of "raw mode tty" and "cbreak mode tty":

raw

BSD cfmakeraw()
Python curses.raw()
stty raw on POSIX, Linux, BSD, and Solaris [ from manpages ]. The one on my Debian Linux system is the GNU coreutils stty(1), which is the most general: it clears all the input flags that are there on the POSIX manpage of stty, which includes IXOFF.
prompt_toolkit's setraw!
the current python tty.setraw()

cbreak

Python curses.cbreak(); based on what this does, and https://pubs.opengroup.org/onlinepubs/7908799/xcurses/intov.html#tag_001_005_002, our mkcbreak() function should also clear ICRNL; however, it should not modify ISIG and IXON.
stty cbreak
the current Python tty.setcbreak()

I would like to put down what I learned about IXOFF today: IXON enables Xon/Xoff "software flow control" on output, and IXOFF controls that on the input. Therefore, enabling or disabling IXOFF has no effect on xterm(1) or the ttys provided by the linux kernel (/dev/tty1, /dev/tty2,...); it is probably more relevant when working with a hardware terminal; with a terminal emulator on a fast pc, that does not matter. See https://tldp.org/HOWTO/Text-Terminal-HOWTO-11.html, https://pubs.opengroup.org/onlinepubs/9699919799/basedefs/V1_chap11.html#tag_11_01_09. Both C-s and C-q are only affected when IXON are turned on or off; if IXANY is also enabled, then any character can be used to resume the flow control as opposed to just C-q (or whatever is the START character).

Edit: @asvetlov Sir, I just realized after reading the devguide that "non-trivial contributions are credited in the Misc/ACKS file" and in the "news entry". Am I eligible for this? Would it be better if I removed my name/email from the main files? I have absolutely no problem in doing that; I just want to understand the correct protocol :)

More info: The "Scroll Lock" key on this 1991 IBM Model M can also be used for "software flow control"; this key is unaffected by either IXON or IXOFF.

old comment 5

New commit: winsize is now a separate class to enforce nonnegative integer entries.

old comment 6

@jonathanslenders I have made the changes that you requested. I request you to please review this again when you have time.

Notes: winsize is now a class so that we can force the fields to be nonnegative. As you had hinted before, the ws_xpixel and ws_ypixel might be less useful now. I am not even sure if there exists an OS that still actually uses them. Should we remove them? They are certainly not hurting other functionality at the moment. The "get size" functionality already exists as os.get_terminal_size(): please see os_get_terminal_size_impl() in https://github.com/python/cpython/blob/master/Modules/posixmodule.c. However, that function also supports windows CONIO; further, there is no function for setting window size. I need both "get" and "set" window size functions (only on *nix) to be able to fix some problems with pty: see https://bugs.python.org/issue41818. Further, note that shutil.get_terminal_size() is not appropriate for this since it simply queries the variables LINES and COLUMNS instead of using TIOCGWINSZ. Thank you for you time.

Edit: to ease the review process, here are the various implementations of login_tty(3):

Edit2: Here is where I am using tty.login_tty() [ it is still named tty.login() there ]: https://github.com/8vasu/pypty2/blob/master/pty2.py. This will be a part of the next (which is possibly the final) PR in this sequence.

Edit3: List of all POSIX.1-2017 input mode flags - https://pubs.opengroup.org/onlinepubs/9699919799/basedefs/termios.h.html#tag_13_72_03_02

old comment 7

TIOCGWINSZ and TIOCSWINSZ are widely available, but they are not POSIX compliant since POSIX generally does not standardize the ioctl() except for the STREAMS interface: see https://www.austingroupbugs.net/view.php?id=1151

They have decided to include functions tcgetwinsize() and tcsetwinsize() in IEEE Std 1003.1 ("POSIX.1") issue 8 [ the upcoming version of POSIX ]; see https://www.austingroupbugs.net/view.php?id=1151#c3856. NetBSD has already implemented them:

Therefore, now I am highly tempted to change the winsize class back to a couple of functions named tcsetwinsize(), tcgetwinsize() OR at least rename the methods to tcsetwinsize(), tcgetwinsize(). I need some help in making this decision. The benefit of having a class is that we can then have named attributes AND enforce nonnegative integer values for them.

Additional notes:

In issue 8, POSIX will also standardize SIGWINCH, which we need for the pty module.
They were originally going to name the functions tcsetsize(), tcgetsize(); see

Edit: renamed the methods; I am keeping the class definition.

old comment 8

@asvetlov @jonathanslenders Thank you for your patience. I have added the final test for pty to this pull request; please look for test_winch(). This test fails when current stdin is a tty and when SIGWINCH is available; it pinpoints the reason why pty.spawn() is unable to dynamically resize the slave window, which leads to gibberish output after, say, the current stdin is resized because of the creation of a tmux pane, or an emacs window in a frame while using ansi-term, or when resizing the GUI [X/Wayland] terminal window. See before and after pictures here: https://bugs.python.org/file49455/before.png, https://bugs.python.org/file49456/after.png

There should only be one more pull request after this :D

Rebased to have only one clean commit. Removed login_tty(); it will be a part of the final pull request. All functions/methods now have standard POSIX/BSD names or are inspired by such names.

… cfmakeecho(), cfmakeraw(), and cfmakecbreak(). The functions setcbreak() and setraw() now return original termios to save an extra tcgetattr() call. Signed-off-by: Soumendra Ganguly <soumendraganguly@gmail.com>

@jonathanslenders Sir, can you please review this again? I have removed a lot of the code [ winsize and login_tty() ]. Thank you in advance.

Signed-off-by: Soumendra Ganguly <soumendraganguly@gmail.com>

This PR is stale because it has been open for 30 days with no activity.

Still waiting for change review.

This PR is stale because it has been open for 30 days with no activity.

@asvetlov @gpshead Please take a look at this if time permits. You can ignore all comments made before since most of that is now covered by termios.tcgetwinsize(), termios.tcsetwinsize(), and os.login_tty().

New summary for this PR: add cfmakeraw(), cfmakeecho(), and cfmakecbreak(); make setraw() and setcbreak() return saved termios attributes to avoid extra tcgetattr() calls.

the-knights-who-say-ni added the CLA signed label Nov 29, 2020

bedevere-bot added the awaiting review label Nov 29, 2020

8vasu mentioned this pull request Nov 29, 2020

bpo-41818: Fix test_master_read() so that it succeeds on all platforms that either raise OSError or return b"" upon reading from master #23536

Merged

asvetlov requested changes Nov 29, 2020

View reviewed changes

bedevere-bot removed the awaiting review label Nov 29, 2020

bedevere-bot added the awaiting changes label Nov 29, 2020

8vasu force-pushed the prepare_tty_for_pty branch from 16a0c70 to ce6c808 Compare November 30, 2020 09:48

bedevere-bot added awaiting change review and removed awaiting changes labels Nov 30, 2020

bedevere-bot requested a review from asvetlov November 30, 2020 10:44

jonathanslenders reviewed Nov 30, 2020

View reviewed changes

8vasu force-pushed the prepare_tty_for_pty branch from 682510f to db202cb Compare December 1, 2020 13:33

8vasu force-pushed the prepare_tty_for_pty branch 7 times, most recently from 6033b85 to a2d2d1d Compare December 5, 2020 08:59

8vasu force-pushed the prepare_tty_for_pty branch 6 times, most recently from 80fbded to a639728 Compare December 7, 2020 18:18

8vasu mentioned this pull request Dec 8, 2020

bpo-41818: Add termios.tcgetwinsize(), termios.tcsetwinsize(). Update docs. #23686

Merged

8vasu marked this pull request as draft December 9, 2020 02:12

8vasu mentioned this pull request Dec 11, 2020

bpo-41818: Add os.login_tty() for *nix. #23740

Closed

Soumendra Ganguly: new additions to the tty library. Functions added:…

a684142

… cfmakeecho(), cfmakeraw(), and cfmakecbreak(). The functions setcbreak() and setraw() now return original termios to save an extra tcgetattr() call. Signed-off-by: Soumendra Ganguly <soumendraganguly@gmail.com>

8vasu force-pushed the prepare_tty_for_pty branch from a639728 to a684142 Compare December 11, 2020 05:25

8vasu marked this pull request as ready for review December 11, 2020 05:36

Clear all POSIX.1-2017 local mode flags in tty.cfmakeraw().

3673a79

Signed-off-by: Soumendra Ganguly <soumendraganguly@gmail.com>

github-actions Bot added the stale Stale PR or inactive for long period of time. label Jan 15, 2021

github-actions Bot removed the stale Stale PR or inactive for long period of time. label Jan 28, 2021

github-actions Bot added the stale Stale PR or inactive for long period of time. label Feb 28, 2021

8vasu mentioned this pull request Nov 20, 2021

bpo-41818: Add os.login_tty() for *nix. #29658

Merged

8vasu mannequin mentioned this pull request May 5, 2022

Lib/pty.py major revision #85984

Open

8vasu and others added 2 commits May 6, 2022 18:37

Remove news file.

f86a5f8

📜🤖 Added by blurb_it.

8288297

ezio-melotti removed the CLA signed label Jul 13, 2022

github-actions Bot removed the stale Stale PR or inactive for long period of time. label Aug 1, 2022

8vasu closed this Feb 11, 2023

bpo-41818: Make additions to the tty module by 8vasu · Pull Request #23546 · python/cpython · GitHub

Repository navigation

bpo-41818: Make additions to the tty module - #23546

bpo-41818: Make additions to the tty module#235468vasu wants to merge 4 commits intopython:mainpython/cpython:mainfrom 8vasu:prepare_tty_for_pty8vasu/cpython:prepare_tty_for_ptyCopy head branch name to clipboard

Conversation

8vasu commented Nov 29, 2020 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Notes

Uh oh!

8vasu commented Nov 29, 2020 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

asvetlov left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

8vasu Nov 29, 2020 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

bedevere-bot commented Nov 29, 2020

Uh oh!

8vasu commented Nov 29, 2020 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

jonathanslenders commented Nov 30, 2020

Uh oh!

8vasu commented Nov 30, 2020 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

bedevere-bot commented Nov 30, 2020

Uh oh!

jonathanslenders commented Nov 30, 2020

Uh oh!

Choose a reason for hiding this comment

Uh oh!

8vasu Nov 30, 2020 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

bpo-41818: Make additions to the tty module#23546
8vasu wants to merge 4 commits into
python:mainfrom
8vasu:prepare_tty_for_pty

8vasu commented Nov 29, 2020 •
edited
Loading

8vasu commented Nov 29, 2020 •
edited
Loading

8vasu Nov 29, 2020 •
edited
Loading

8vasu commented Nov 29, 2020 •
edited
Loading

8vasu commented Nov 30, 2020 •
edited
Loading

8vasu Nov 30, 2020 •
edited
Loading

8vasu Nov 30, 2020 •
edited
Loading

8vasu commented Dec 1, 2020 •
edited
Loading

8vasu commented Dec 3, 2020 •
edited
Loading