[Scons-users] Fwd:Re: timing issues and protecting from them

[Tom Tanner]

On 9/12/15 21:36, Dirk Bächle wrote:
> Tom,
>
> thank you very much for your clarifications. If I read and understood 
> your additional comments correctly, you don't think that your patch 
> proposal will allow "modification of source files outside of SCons 
> during a complete build" in general, right?
> Because, that's what I (and Bill and William too, probably) had 
> assumed at first.

Umm. I don't think so. it's more spotting that a file has changed in 
such a way that could cause an individual target to have the wrong 
information and stopping you getting wrong information in the cache or 
in your build.
>
> You want to extend the "source files may change" period up to the 
> point in time where the build of a single target has finished. I can 
> see that as an additional "--paranoid-built-checking" option, as Bill 
> suggested. So if you want to prepare a pull request for this, go right 
> ahead.

Well, my original question was - where would be a good place to do this, 
which hasn't really been answered.

> But allow me a more practical question: Let's assume that I have a C++ 
> build with about 1000 files of the same size and roughly same build 
> time each. Why should I switch on your "--paranoid" option, and how 
> much more safety/stability does it give me for my whole build process?

My experience with not having it is that sometimes you can change a file 
while the build is going on and you end up with scons thinking it 
doesn't need to build a file but in fact it has built with a previous 
version of the source. And it's a nightmare spotting and fixing that. 
Well, fixing it generally involves touching it so the timestamp is 
sufficiently drifted for a recalculation.

My particular worry is I don't know how much it could affect what is in 
your build cache. Having a build cache takes our build time down from a 
day to about half an hour. So having it correct is really rather 
important. And I'd rather be paranoid than have wrong software in 
production, especially as I don't think a quick check of the timestamps 
should be terribly expensive.

>
> Best regards,
>
> Dirk
>
> _______________________________________________
> Scons-users mailing list
> Scons-users at scons.org
> https://pairlist4.pair.net/mailman/listinfo/scons-users