verne.freenode.net changed the topic of #mlpack to: http://www.mlpack.org/ -- We don't respond instantly... but we will respond. Give it a few minutes. Or hours. -- Channel logs: http://www.mlpack.org/irc/
pantsforbirds has joined #mlpack
pantsforbirds has quit [Client Quit]
pantsforbirds has joined #mlpack
pantsforbirds has quit [Remote host closed the connection]
nilay has joined #mlpack
Mathnerd314 has quit [Ping timeout: 244 seconds]
nilay has quit [Ping timeout: 250 seconds]
mentekid has joined #mlpack
nilay has joined #mlpack
marcosirc has joined #mlpack
< rcurtin>
mentekid: for PR #729, in README.md, shouldn't I be able to say that getting the files from Boost 1.58 or newer would work to remove the dependence on Boost>?
< rcurtin>
or is there some specific functionality that 1.61 provides that 1.58 doesn't?
< mentekid>
rcurtin: Let me check
< mentekid>
I think 1.58 adds trigamma which is the latest feature of boost we need
< mentekid>
yep it was added in 1.58... So there shouldn't be a requirement for 1.61
< mentekid>
But I have taken the files from 1.61, so the newest version... There shouldn't be a problem with that though
< rcurtin>
yeah; I just wanted to update the README to have the minimum required version of boost for someone who wants to strip the code from our release
< rcurtin>
in case they have some problem with the Boost license or something ...
< mentekid>
yes I was still a bit lost when I wrote the README so it was a mistake
< rcurtin>
it's hard for me to see how such a situation would arise, but hey, I figure it's helpful to provide documentation for how to remove our backports if needed
< mentekid>
I guess at some point we might end up using so many features of newer versions we might need to increase the complete dependency and remove all the backports
< mentekid>
I've pushed the changes, I mention that I took the files from Boost 1.61.0 but that Boost 1.58 is the minimum requirement for removing the backport
< rcurtin>
I agree, at some point it will become reasonable to upgrade the minimum required version
< rcurtin>
but generally my threshold for what the maximum minimum version can be is set by whatever the oldest still-in-wide-usage version of RHEL provides
< rcurtin>
(which is why we are still requiring only CMake 2.8 not CMake 3.1 or so)
< mentekid>
I think especially if it's small features (like trigamma here) it's not nice to inconvenience a bunch of users and make them install new versions of our dependencies
< rcurtin>
yeah, I agree
< mentekid>
on the flipside now my main contribution (in loc) for GSoC will be code I didn't write :P
< rcurtin>
the most recent time we changed the dependency version was to get the boost priority_queue object for the cosine tree
< rcurtin>
hah :)
< mentekid>
rcurtin: I broke the build
< mentekid>
I think it's just the off-by-one in boost_backport.hpp line 23
< mentekid>
I checked BOOST_VERSION < 105500, but this way I missed 105500 which is where the build failed
< rcurtin>
ah, ok, if you want to submit a PR I'll merge it
< rcurtin>
hm, hang on, 105500 is what was used in the old prereqs.hpp... why do you say the build is broken?
< rcurtin>
it looks like Travis hasn't finished yet
< rcurtin>
also, I made a few changes to the code, to try and set it up better for when the rest of the distribution support is added
< rcurtin>
the main interesting change is f364cba, which changes Alpha() and Beta() to Alpha(dim) and Beta(dim), and adds a Dimensionality() method to get the number of dimensions in the distribution
< rcurtin>
let me know if I broke anything, hopefully there is no issue :)
< mentekid>
I just got an email from Jenkins telling me the build with 1.55 failed
< mentekid>
it was sent to me and you (at gth671b@mail.gatech.edu)
< mentekid>
Why did we check if boost < 105500, since 105500 doesn't have unordered_map.hpp?
travis-ci has joined #mlpack
< travis-ci>
mlpack/mlpack#1275 (master - 2c026d8 : Ryan Curtin): The build was fixed.
< rcurtin>
ah, gth671b@mail.gatech.edu is dead... let me update that email then
< rcurtin>
no wonder Jenkins has been so quiet lately :)
< rcurtin>
oh! I think I see what is happening here. in previous revision, there was an error: BOOST_VERSION was never defined, so the backported version was always used
< rcurtin>
in the code you committed, this is fixed, so the "105500" was wrong all along
< mentekid>
ah right
< mentekid>
two wrongs do some time make a right
< rcurtin>
:)
< rcurtin>
I've reconstructed my history here, it looks like I even checked that boost 1.55.0 did not have unordered_map.hpp and boost 1.56.0 did, but then I wrote down the wrong version number
< rcurtin>
or technically the wrong inequality--- "<=" would have worked
chrismeyer has joined #mlpack
sumedhghaisas has joined #mlpack
travis-ci has joined #mlpack
< travis-ci>
mlpack/mlpack#1280 (master - 3c119a3 : Ryan Curtin): The build was fixed.