[Scons-users] Decider which checks md5 checksums unexpectedly detecting changed/invalid ones

Tom Tanner (BLOOMBERG/ LONDON) ttanner2 at bloomberg.net
Thu Sep 26 10:46:39 EDT 2013


No problems. I seem to be asking a lot of odd questions at the moment.

----- Original Message -----
From: garyo at oberbrunner.com
To: Tom Tanner (BLOOMBERG/ LONDON), scons-users at scons.org
At: Sep 25 2013 14:48:59

Hi Tom -- I'm not ignoring you, but the issues you raise require some
thought. Just wanted to let you know.


On Mon, Sep 23, 2013 at 5:04 AM, Tom Tanner (BLOOMBERG/ LONDON) <
ttanner2 at bloomberg.net> wrote:


> Replying to myself again:

>

> The problem appears to be to do with copying something from the cache: If

> you copy a file from the cache, the timestamp can go backwards rather than

> forwards. So the '(time.time() - mtime) > max_drift' test can spuriously

> return true (especially if max_drift is quite low because you have a stable

> clock over your network, as recommended by the 'GoFastButton' on the wiki).

>

> I am at a bit of a loss here, as what the wiki seems to say (only cache

> the signature if) and what the code seem to do (use the cached signature

> if) appear to be rather different.

>

> The option appears to be to either (1) set the max_drift to negative or

> (2) to force the sig to be recalculated after a build in the built()

> function in Node/__init__.py. You certainly need to do the later if you've

> got the max drift set to 0, or you'll get wrong builds. But I'm not sure

> what get_max_drift_csig is actually meant to be doing.

>

> Help would be appreciated as I want my build to be both fast an accurate.

>

> _______________________________________________

> Scons-users mailing list

> Scons-users at scons.org

> http://four.pairlist.net/mailman/listinfo/scons-users

>

>



--
Gary



More information about the Scons-users mailing list