Update the boost version to 1.62

[greglandrum]

Are we constrained to stick with boost 1.56 or could we go ahead and update to the most recent version?

[rvianello]

in general, I think we would mainly need to check if we are able to build rdkit with boost 1.62 on centos6

I am not aware about the specific constraints, but I think @bp-kelley tested several versions of boost recently and suggested using 1.58, so he might have some more concrete suggestions.

[greglandrum]

I can do the centos6 test; I have a VM that I use for my conda testing.
I'll give that a try this evening.

[greglandrum]

I wonder about VS2008? (Or maybe we should explore moving to something "a bit" more up-to-date for py27 on windows)

[greglandrum]

I updated the recipe for boost and the rdkit to use boost=1.62 and the rdkit release branch (just to ensure up-to-date production code for the testing phase) and it builds fine on centos6 with the default compilers.

I'll push a branch and give it a try on the mac and windows too.
And, of course, things don't "just work" on the Mac. I'll have to explore a bit more.

[greglandrum]

the other thing that is worth considering, and that I'm trying now as well, would be to use the normal conda version of boost: 1.61. This would spare us from having to worry about it at all.

[rvianello]

boost 1.61 I'm afraid would still cause problems on the recent linux distros, unless a fix for boost::multiprecision was backported into the anaconda package. I was testing a build using the packages from conda-forge (that provides boost 1.62, and a lot of other updates), but I'm still troubleshooting some small issues.

regarding py27 on windows, if we move away from vc2008, then I guess we can just use vc2015, I'm not sure using any intermediate release would be different in terms of portability, and I think most (if not all) of the earlier vc releases are no longer available for download.

[greglandrum]

But 1.61 would just cause problems for people who want to build their own version of the RDKit using the conda recipes, right? That's not a huge concern to me.

[greglandrum]

v1.61 (the conda version) works fine on:

• windows (using vc2015)
• centos6
• Mac

[rvianello]

ah, yes, streamlining the overall process and reducing the number of custom packages would surely help (I initially misunderstood the intention could be related to the bug in boost < 1.62)

[greglandrum]

Na, the intention is just to simplify. :-)

[greglandrum]

v1.61 also seems to be somewhat problematic with gcc5.x (Ubuntu 16.04). The regex problem with the SLN library appears there

[greglandrum]

but I didn't yet try the conda version. I assume that has the same problem

[bp-kelley]

I seem to remember that 1.61 and osx was problematic as well.

…

On Thu, Dec 22, 2016 at 10:42 AM, Greg Landrum ***@***.***> wrote: but I didn't yet try the conda version. I assume that has the same problem — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#42 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AJbioP06f1XQTRqAVCd7duYLsx9dPvpKks5rKpp3gaJpZM4LTA-v> .

[bp-kelley]

Looks like boost 1.62 just hit Travis OS X. I'll update my build matrix and try a new PR https://blog.travis-ci.com/2016-12-13-xcode-82-ga-is-here

…

---- Brian Kelley

On Dec 22, 2016, at 11:08 AM, Brian Kelley ***@***.***> wrote: I seem to remember that 1.61 and osx was problematic as well. > On Thu, Dec 22, 2016 at 10:42 AM, Greg Landrum ***@***.***> wrote: > but I didn't yet try the conda version. I assume that has the same problem > > — > You are receiving this because you were mentioned. > Reply to this email directly, view it on GitHub, or mute the thread. >

[bp-kelley]

I take it back, it's home-brew boost and python 27 only. Still need to install miniconda. Cheers, Brian

…

On Fri, Dec 23, 2016 at 8:01 AM, Brian Kelley ***@***.***> wrote: Looks like boost 1.62 just hit Travis OS X. I'll update my build matrix and try a new PR https://blog.travis-ci.com/2016-12-13-xcode-82-ga-is-here ---- Brian Kelley On Dec 22, 2016, at 11:08 AM, Brian Kelley ***@***.***> wrote: I seem to remember that 1.61 and osx was problematic as well. On Thu, Dec 22, 2016 at 10:42 AM, Greg Landrum ***@***.***> wrote: > but I didn't yet try the conda version. I assume that has the same problem > > — > You are receiving this because you were mentioned. > Reply to this email directly, view it on GitHub > <#42 (comment)>, > or mute the thread > <https://github.com/notifications/unsubscribe-auth/AJbioP06f1XQTRqAVCd7duYLsx9dPvpKks5rKpp3gaJpZM4LTA-v> > . >