Definitely. I check out the libraries, mostly because I am frequently offline. I do actually use a caching server for Python dependencies, actually, for Python development--and it works quite well, but the audience for that program is programmers, not hardware designers.
Adam Wolf On Wed, Jan 13, 2016 at 12:07 PM, Wayne Stambaugh <[email protected]> wrote: > On 1/13/2016 12:20 PM, Adam Wolf wrote: > > I can test this on OS X this week. > > Thanks. > > > > > (I am excited for Jon's git plugin roadmap. If Github works well for > > other people, awesome, but I am not a fan of depending on proprietary > > software, and based on the chatter on the email list and Dick's > > suggestion that users install and configure a webserver/caching > > proxy/modify /etc/hosts suggests to me we have room for improvement.) > > Please keep in mind, nginx is just a suggestion not a demand. This is > how Dick resolved his github performance issues. It is not a shooting > offense. I would suggest that if your not interested in using some type > of caching proxy such as nginx that you clone the footprint repos to > your hard drive and configure your fp-lib-table accordingly. No one is > being force to use anything they are not comfortable using. Without > some type of caching, the github plugin will only ever be as fast as > your internet connection no matter how well we optimize the code. > > A git plugin could be interesting. We'll see how it pans out in > practice. The only thorny issue may be using the git library depending > on the platform. > > > > > Adam Wolf > > Cofounder and Engineer > > W&L > > > > On Wed, Jan 13, 2016 at 11:01 AM, Wayne Stambaugh > > <[email protected] <mailto:[email protected]>> wrote: > > > > Would someone please test this patch on osx and on a linux distro > that > > uses a libcurl variant built against ntls or gnutls. I will test it > on > > windows. Please read the comments below carefully and do not reply > to > > Dick directly. > > > > -------- Forwarded Message -------- > > Subject: [PATCH] github_plugin -> curl -> openssl thread safety > > Date: Wed, 13 Jan 2016 00:07:35 -0600 > > From: Dick Hollenbeck <[email protected] <mailto:[email protected]>> > > To: Wayne Stambaugh <[email protected] > > <mailto:[email protected]>>, [email protected] > > <mailto:[email protected]> > > <[email protected] <mailto:[email protected]>> > > > > Guys, this brings in the necessary locks required to make openssl > thread > > safe. I think it completes the migration to libcurl. As I reported > to > > Wayne, even on linux I was seeing evidence of unprotected critical > > sections of code. Going down to 6 threads merely swept the problem > > under a rug but did not fix the fundamental problem. This does. > > Six threads still seems about the optimal mix of chaos and > throughput. > > > > You may want to bring back the FindOpenSSL.cmake script which this > must > > use, but I got by with CMake's default. Please just do that as a > > separate patch since I am not responsible for it. You could also do > any > > Windows specific fixups as a separate patch also, as it does not > > attribute stuff to me that I did not do, and vice versa. > > > > HTH in fulfilling your vision for the github plugin. The locking > does > > not affect the speed by much, and multiple threads are still a > > tremendous speed boost. I still would never use the plugin without > > nginx, and I think that should be made more clear to those who wine > > about speed. There is an awful lot of whining, and none of these > people > > even seem to know about nginx. nginx + github plugin is even faster > > than the PCB_IO plugin. > > > > I think I am loading about 80 libraries in 2 seconds. Only two of > those > > are PCB_IO, the rest are github. > > > > PLEASE MAKE THIS KNOWN! > > > > > > *) Switch to static linking of libcurl and on linux and windows and > also > > statically link in only required portions of openssl. > > *) Add the required thread locks which openssl needs. > > *) Remove the get curl version call from BASEFRAME since it pulls in > > curl and openssl > > into every derived wxFrame class link image. > > *) Remove curl function from PGM_BASE, switch to atexit() instead. > > Anything in PGM_BASE made the singletops bigger. > > *) Tested on Linux only. > > > > > > _______________________________________________ > > Mailing list: https://launchpad.net/~kicad-developers > > Post to : [email protected] > > <mailto:[email protected]> > > Unsubscribe : https://launchpad.net/~kicad-developers > > More help : https://help.launchpad.net/ListHelp > > > > > >
_______________________________________________ Mailing list: https://launchpad.net/~kicad-developers Post to : [email protected] Unsubscribe : https://launchpad.net/~kicad-developers More help : https://help.launchpad.net/ListHelp
