Subject: Re: [boost] El Capitan issues (Was: [1.61] Two weeks remaining for new libraries and breaking changes)
From: Robert Ramey (ramey_at_[hidden])
Date: 2016-02-15 11:19:09
On 2/14/16 11:58 PM, Vladimir Prus wrote:
>> a) Current status is that I can pass all my tests on my local machine
>> - Mac OSX when building static libraries. But now
>> - due to a change from an "upgrade" to El Capitan - boost build won't
>> build and run tests for shared libraries on this
> It's certainly unfortunate that Apple chose to break some many things in
> El Capitan,
Of course I would like to see this fixed - but truth is I don't think
it's holding me up all that much. My normal mode is to test with
static libraries and then check in to the develop branch and watch the
test matrix. These days it never happens that it passes on the static
library test but fails on the shared library tests. So as I said it's
not holding me up that much.
It is of value to run in all 4 combinations of static/shared
debug/release because some combinations eliminated code which is not
explicitly called - so called dead code. But the serialization library
can require code which is not explicitly referred to by name - code
which "looks like" dead code. So one has to take special precautions to
make sure that code doesn't get eliminated. The good news is that once
one has done this, it seems the problem is eliminated once and for all -
at least for the last 10 years.
> and it's alarming that you're the first to notice the issue as it related
> to Boost testing.
Well, El Capitan has been out only a short time - and I'm sure a lot of
people have not upgraded so I wouldn't get alarmed.
> discussed some options on Boost.Build list,
> and hopefully we'll have them implemented in time for 1.61.
To me, more urgent is the problem associated with b2 on cygwin. I've
noted on this list. Basically if one builds b2 using the bootstrap.sh
under cygwin it seems that it get's the / and \ mixed up in the
pathnames. And this prevents b2 from working correctly. I wouldn't be
surprised if this isn't also a problem with mingw on windows as well.
I'm guessing this probably an easy fix for you.
I also mentioned that the "regression" tools aren't tested in accordance
with Boost customs. I think you (or whoever) should consider this.
These tools only are on the develop but not the master branch - and as
far as I know - aren't subjected to regular boost testing. I know it
seems like a pain to setup - but once it's been done you get a lot of
testing without having to spend any time on it.
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk