Openembedded Core Discussions
 help / color / mirror / Atom feed
From: Richard Purdie <richard.purdie@linuxfoundation.org>
To: michael.haener@siemens.com, openembedded-core@lists.openembedded.org
Cc: Adrian Freihofer <adrian.freihofer@siemens.com>,
	Peter Marko <peter.marko@siemens.com>
Subject: Re: [OE-core] [PATCH] sstate: allow marking a cached object as in use without touching mtime
Date: Wed, 09 Sep 2026 23:40:46 +0100	[thread overview]
Message-ID: <cb88917248190287dcd8ec498180258f2dcb7a5c.camel@linuxfoundation.org> (raw)
In-Reply-To: <20260909212205.50713-1-michael.haener@siemens.com>

On Wed, 2026-09-09 at 23:22 +0200, Michael Haener via lists.openembedded.org wrote:
> Every access to an object in the sstate cache is recorded with a plain
> touch, which writes both the modification time and the access time.
> 
> Add SSTATE_ATIME_UPDATE_AFTER. When it is set, the modification time is no
> longer written, and the access time is refreshed only once the existing
> one is at least that many seconds old.
> 
> This reduces the number of write cycles a build causes on the cache, which
> matters most where many build servers share one sstate cache: each of them
> would otherwise refresh the access time of every object it looks at.
> 
> With mtime and atime kept apart, a cache can also be measured for reuse
> and age.
> 
> When the variable is unset the behaviour is unchanged.
> 
> Signed-off-by: Michael Haener <michael.haener@siemens.com>
> Reviewed-by: Adrian Freihofer <adrian.freihofer@siemens.com>
> Reviewed-by: Peter Marko <peter.marko@siemens.com>
> ---
>  meta/classes-global/sstate.bbclass | 59 ++++++++++++++++++++++--------
>  1 file changed, 43 insertions(+), 16 deletions(-)

Whilst I can understand the motivation, I can't say I'm thrilled by
complicating this code again. I'd recently tried to simplify it as the
different access patterns were causing various problems in their own
right. This adds a new variable, two implementations (shell and python)
of an algorithm and it just makes the code more complex and hard to
understand.

Cheers,

Richard

 


      reply	other threads:[~2026-09-09 22:40 UTC|newest]

Thread overview: 2+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-09-09 21:22 [PATCH] sstate: allow marking a cached object as in use without touching mtime Michael Haener
2026-09-09 22:40 ` Richard Purdie [this message]

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=cb88917248190287dcd8ec498180258f2dcb7a5c.camel@linuxfoundation.org \
    --to=richard.purdie@linuxfoundation.org \
    --cc=adrian.freihofer@siemens.com \
    --cc=michael.haener@siemens.com \
    --cc=openembedded-core@lists.openembedded.org \
    --cc=peter.marko@siemens.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox