Fix bad exceptions handlings

[4383]

This patch fix 2 conditions that are badly designed. They can't work. In all case the exception will
be catched by the first except and then the second one will be all the time ignored.

Here is a tiny PoC to demonstrate the bad design:

import errno

def trigger(err, message):
    print(f"try to simulate {message}: {err}")
    raise OSError(err, message)


def catcher(err, message):
    try:
        trigger(err, message)
    except OSError as e:
        print("I'm here")
        if e.errno != errno.EAGAIN:
            raise
    except OSError as e:
        print("I'm there")
        if e.errno == errno.EPIPE:
            return ''
        raise


if __name__ == "__main__":
    try:
        catcher(errno.EPIPE, "pipe!")
    except Exception as e:
        print(e)
    try:
        catcher(errno.EAGAIN, "again!")
    except Exception as e:
        print(e)

The previous code will output:

try to simulate pipe!: 32                                                      
I'm here                                                                       
[Errno 32] pipe!                                                                                                                                               
try to simulate again!: 11     
I'm here   

[codecov]

Codecov Report

Attention: 3 lines in your changes are missing coverage. Please review.

Comparison is base (4cf3190) 55% compared to head (eed4bb0) 55%.

Files	Patch %	Lines
eventlet/green/os.py	0%	3 Missing ⚠️

Additional details and impacted files

@@          Coverage Diff          @@
##           master   #883   +/-   ##
=====================================
  Coverage      55%    55%           
=====================================
  Files          89     89           
  Lines        9837   9832    -5     
  Branches     1828   1825    -3     
=====================================
  Hits         5414   5414           
+ Misses       4054   4049    -5     
  Partials      369    369           

Flag	Coverage Δ
ipv6	22% <0%> (+<1%)	⬆️
py310asyncio	51% <0%> (+<1%)	⬆️
py310epolls	52% <0%> (+<1%)	⬆️
py310poll	52% <0%> (+<1%)	⬆️
py310selects	52% <0%> (+<1%)	⬆️
py311asyncio	51% <0%> (+<1%)	⬆️
py311epolls	52% <0%> (+<1%)	⬆️
py312asyncio	49% <0%> (-1%)	⬇️
py312epolls	50% <0%> (+<1%)	⬆️
py37epolls	50% <0%> (+<1%)	⬆️
py38epolls	52% <0%> (+<1%)	⬆️
py38openssl	51% <0%> (+<1%)	⬆️
py38poll	52% <0%> (+<1%)	⬆️
py38selects	52% <0%> (+<1%)	⬆️
py39dnspython1	51% <0%> (+<1%)	⬆️
py39epolls	52% <0%> (+<1%)	⬆️
py39poll	52% <0%> (+<1%)	⬆️
py39selects	52% <0%> (+<1%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[itamarst]

Good catch! I wonder if this is the reason for the some of the weird TLS shutdown edge cases I've seen.

[4383]

Good catch! I wonder if this is the reason for the some of the weird TLS shutdown edge cases I've seen.

No idea, my discover is pure serendipity...