Re: [PATCH 21.5] TLS support

Mats Lidell writes:
 > >>>>> Stephen J Turnbull <[email protected]> writes:
 > 
 > > That's the theory.  Amusingly enough, in this very thread we're
 > > observing that Jerry's changes have broken things even though
 > > they're turned off.  That happens to ordinary mortals and Ben Wing
 > > alike.
 > 
 > Well, you won't notice unless it breaks somethings which actually is
 > one of the points. You have to, and get to, fix those integration
 > errors upfront.

In an early stage, it may get fixed by accident as development
proceeds, and never been seen by users at all.  If not, if this
development is done on a branch, you'll get a spate of this kind of
error at integration time.  If you do it on the trunk with ifdefs, not
only are they spread out over time, but there are probably more of
them because ifdefs are fiddly and prone to this kind of issue.

 > Besides beta testers should expect to get some build and runtime
 > errors. That is one reason they are beta testers. To help find
 > problems by tripping over them.

Not really.  They're suppose to find problems that come up in real use
cases that the developers hadn't thought of.  If there is a way for us
to catch bugs before the beta testers get the code, we should use it,
IMHO.

 > But I don't mean like [USE flags] in Gentoo...

 > As you suggest it will probably suffice for us. However the feature
 > flag concept, as I get it, is more of having yet an interface and
 > mechanism for controlling a feature of the system. That mechanism is
 > solely a developer thing and will vary over time depending on what
 > features are being developed.

That's not very helpful; except for "solely a developer thing", it's a
description of USE flags, ifdefs, customize, behaviors, configure, ....

Steve