DMARC-Filter: OpenDMARC Filter v1.4.2 delorie.com 50JH5RKx1799918
DMARC-Filter: OpenDMARC Filter v1.4.2 sourceware.org E56713858D21
Message-ID: <ad851553-ab4c-4438-9cd1-86ab56f01a8a@SystematicSW.ab.ca>
Date: Sun, 19 Jan 2025 10:04:20 -0700
MIME-Version: 1.0
User-Agent: Mozilla Thunderbird
Subject: Re: [Improvement] Cache user auth and converted paths for
 child-processes
Content-Language: en-CA
To: cygwin AT cygwin DOT com
References: <CANM0SV2VaAsQbtwmmwmBmZjbWwnEdvVKGcZSHR02uTow1yV=ig AT mail DOT gmail DOT com>
Organization: Systematic Software
In-Reply-To: <CANM0SV2VaAsQbtwmmwmBmZjbWwnEdvVKGcZSHR02uTow1yV=ig@mail.gmail.com>
Precedence: list
From: Brian Inglis via Cygwin <cygwin AT cygwin DOT com>
Reply-To: cygwin AT cygwin DOT com
Cc: Brian Inglis <Brian DOT Inglis AT SystematicSW DOT ab DOT ca>
Content-Type: text/plain; charset="utf-8"; Format="flowed"
Errors-To: cygwin-bounces~archive-cygwin=delorie DOT com AT cygwin DOT com
Sender: "Cygwin" <cygwin-bounces~archive-cygwin=delorie DOT com AT cygwin DOT com>
Content-Transfer-Encoding: 8bit

On 2025-01-19 05:56, Devste Devste via Cygwin wrote:
> strace -o strace.log dirname -- /some/path/here
> There are 2 points that could make it significantly faster and
> shouldn't be too hard to implement?

Criticisms as expected to be accompanied by code and all Patches are 
Thoughtfully Considered (PTC - see https://cygwin.com/acronyms/) - git 
format-patch/send-email against repo main branch to cygwin-patches@... ML.

All contributors are volunteers working when they have spare time available.

> 1) A significant amount of time is spent on user auth (as seen in
> various github issues the infamous /etc/passwd nsswitch.conf fix)
> Wouldn't it be possible to just reuse the auth data from the current
> shell for the subshell, e.g. for
> basename -- $(dirname -- /some/path/here)

Some of this is done for each child in a Cygwin process tree, and for all Cygwin 
processes if you run cygserver.
You can also decide to limit what sources should be used in nsswitch.conf.

> 2) why are unix/dos path conversions of environment variables not
> cached? A significant amount of time (15-30%, depending on the number
> of environment variables) is spent on the conversion for every
> invocation.
> However, this would be extremely simple to cache and reusable even on
> completely unrelated subshells.
> cache key = original path
> cache value = converted value
> e.g.
> 
> ```
>     44   12839 [main] dirname 16929 mount_info::conv_to_posix_path:
> conv_to_posix_path (C:\Users\User123\bin, 0x10000100, no-add-slash)
>     44   12883 [main] dirname 16929 normalize_win32_path:
> C:\Users\User123\bin = normalize_win32_path (C:\Users\User123\bin)
>     44   12927 [main] dirname 16929 mount_info::conv_to_posix_path:
> mount[0] .. checking / -> C:\Program Files\Git
>     44   12971 [main] dirname 16929 mount_info::conv_to_posix_path:
> mount[1] .. checking /bin -> C:\Program Files\Git\usr\bin
>     44   13015 [main] dirname 16929 mount_info::conv_to_posix_path:
> mount[2] .. checking /tmp -> C:\Users\User123\AppData\Local\Temp
>     44   13059 [main] dirname 16929 mount_info::conv_to_posix_path:
> /c/Users/User123/bin = conv_to_posix_path (C:\Users\User123\bin)
> ```
> could be cached with key
> C:\Users\User123\bin
> and value
> /c/Users/User123/bin
> at least for the current process (e.g. a bash script and it's
> children) without risking any noticeable outdated cache issues
> (probably even longer, however we want to keep it simple and don't
> want to worry about cache invalidation too much)

Any process can change the environment, so contents can not be assumed, and 
everything has to be rechecked each time.
It is probably quicker (44µs) to check for paths and convert.

Drop all the Windows paths and variables from your Cygwin environment, unless 
you intend to run Windows programs from there, and you will not have so much 
overhead.

For a cache, you would want to key from the env var name, have a path flag, and 
keep Windows and Cygwin alternatives, but that only helps if you have child 
processes where the environment stays the same, so the penalty should be paid in 
any children.

-- 
Take care. Thanks, Brian Inglis              Calgary, Alberta, Canada

La perfection est atteinte                   Perfection is achieved
non pas lorsqu'il n'y a plus rien à ajouter  not when there is no more to add
mais lorsqu'il n'y a plus rien à retrancher  but when there is no more to cut
                                 -- Antoine de Saint-Exupéry

-- 
Problem reports:      https://cygwin.com/problems.html
FAQ:                  https://cygwin.com/faq/
Documentation:        https://cygwin.com/docs.html
Unsubscribe info:     https://cygwin.com/ml/#unsubscribe-simple