kicad-developers team mailing list archive

[Collin Anderson <metacollin@xxxxxxxxxxxx>]

Sorry, I didn't mean to waste anyone's time with this patch, I am still very new to the project and still getting a feel for where things are heading and what is a dead end and what isn't.  Additionally, I would be terrified of just jumping versions  of such a key component like boost without a lot of testing first.  Boost 1.54 unfortunately is incompatible with OS 10.10, that's ultimately the motivation behind this patch. 

As for ripping all that crap out, that was exactly what I wanted to do too.  Originally, my homebrew tap had been using the KICAD_SKIP_BOOST option and using an actual install of regular boost 1.57, and it seemed to work without issue for a wide variety of machines, versions of OS X, and users.  Beyond that, all but one of the fixes had been pulled into 1.57.  The missing patch curiously was simply commenting out two useless lines in minkowski.hpp, I assumed half out of removing warnings, and half out of sheer principal, lines of code with no purpose should not clutter up source files, in my opinion.

However, I discovered, using my own as well as reports from several other people, that the autoroute mode in pcbnew crashes the entire .app almost instantly, and with total consistency.  Commenting out those useless lines fix this issue completely.  I'm not sure if I discovered anything, or simply rediscovered for myself something well known by other contributors.  When you say next stable release, do you mean of KiCad, or of boost? If you mean to say that the 1.58 release of boost will incorporate the minkowski patch, then terrific, I can't wait for all that crap to get ripped out.  It proved to be a *huge* headache when making the homebrew tap, thats for sure!

Or did you mean next stable release of KiCad? Sorry, I am honestly not sure.  

Assuming this was all well known to everyone, I was curious if anyone knew the mechanism behind the minkowski autoroute issue? My quick guess is that the unused variables being set equal to vital pointers in the convolve function is causing a compiler optimization to do something unfortunate to those poor pointers, and loss of whatever memory location they were supposed to be holding.  *shrug*


-- 
"Violence is the last refuge of the incompetent." - Isaac Asimov

> On Feb 17, 2015, at 2:34 PM, Wayne Stambaugh <stambaughw@xxxxxxxxx> wrote:
> 
> Review: Disapprove
> 
> After the next stable release, this patch will be lost.  I plan on ripping out all of the custom dependency download, patch, and build code from the CMake files after the next stable.  If someone needs this to build dependencies on their platform than it will have to be maintained as an external project.  The complexity it has added to the project configure and build has been a major source of pain.  I recommend creating a kicad-boost project that downloads, patches, and builds boost for the platforms that need it.
> -- 
> https://code.launchpad.net/~metacollin/kicad/boosting_boost/+merge/248085
> You are the owner of lp:~metacollin/kicad/boosting_boost.