[Scons-users] Scons none-deterministic behavior for incremental build

[Hua Yanghao]

On Wed, Dec 5, 2018 at 5:27 PM Mats Wichmann <mats at wichmann.us> wrote:
>
> On 12/5/18 8:34 AM, Bill Deegan wrote:
> > Indeed getting a complicated build to work with any build system is
> > non-trivial.
> >
> > -Bill
>
> I don't want to wander too far off into philosophical arguments (oh, who
> am I kidding!), so I'll drop this thread after making this observation:
> as a user who wandered over here after not being clear how to
> disentangle problems in a built that had chosen scons, it seems to me
> way too easy to get things wrong in scons. With great flexibility comes
> great opportunities to mess up. Or something like that :)

@Mats this is exactly my observations too. I am not blaming scons at
all -- I love scons -- I am just trying to say scons in my humble
opinion is positioned at a lower level that have provided a good
foundation for any particular "pattern" to be implemented on top.
Those "pattern" obviously is not of scons concern and there are too
many of them (kbuild being one of the classic patterns), and implement
such a "pattern" is not very easy in scons -- arguably even harder in
makefiles, just take a look at kbuild source code, a mixed shell
scripts & C code -- so there should be room for "patterns" implemented
on top of scons to make a particular class of build way easier to use,
and in my case this is for cross-compiling embedded C program.

My new scons-based build system since give it to users I hear not a
single complain because it is so simple it is very difficult to get it
wrong. We tried many years back to ask users to use plain scons, we
didn't go anywhere but kept using makefiles because people are already
familiar with it.