Re: [sage-devel] make behaviour

2014-11-04 Thread Volker Braun
On Tuesday, November 4, 2014 12:18:22 PM UTC, Jeroen Demeyer wrote: > > So is it comparable to ccache then (caching all build output, not just > from .c files)? Yes except that ccache doesn't cache the building, tuning, and linking... -- You received this message because you are subscribed to

Re: [sage-devel] make behaviour

2014-11-04 Thread Jeroen Demeyer
On 2014-11-03 10:23, Ralf Stephan wrote: Thanks, I should use sage -f more often. So, what became of Jeroen's idea of using the "order only dependency" feature of GNU make? I created http://trac.sagemath.org/ticket/17286 -- You received this message because you are subscribed to the Google G

Re: [sage-devel] make behaviour

2014-11-04 Thread Jeroen Demeyer
On 2014-11-03 12:15, Volker Braun wrote: On Monday, November 3, 2014 9:44:10 AM UTC, Jeroen Demeyer wrote: I don't know about hashdist, but I don't see how it could magically fix the "compile time when switching branches" issue. By caching properly, and using tarball + build script + v

Re: [sage-devel] make behaviour

2014-11-03 Thread Volker Braun
On Monday, November 3, 2014 9:44:10 AM UTC, Jeroen Demeyer wrote: > > I don't know about hashdist, but I don't see how it could magically fix > the "compile time when switching branches" issue. > By caching properly, and using tarball + build script + variables as hash key. So you never have to

Re: [sage-devel] make behaviour

2014-11-03 Thread Jeroen Demeyer
On 2014-11-03 10:37, Volker Braun wrote: A better long-term plan is IMHO to use hashdist, which solves both the "compile time when switching branches" and LD_LIBRARY_PATH issues. I don't know about hashdist, but I don't see how it could magically fix the "compile time when switching branches" is

Re: [sage-devel] make behaviour

2014-11-03 Thread Volker Braun
This probably needs a bunch of testing to figure out which dependencies are not really compiled-in, so I wouldn't want to rush it into 6.4. A better long-term plan is IMHO to use hashdist, which solves both the "compile time when switching branches" and LD_LIBRARY_PATH issues. On Monday, Nov

Re: [sage-devel] make behaviour

2014-11-03 Thread Jeroen Demeyer
On 2014-11-03 10:23, Ralf Stephan wrote: Thanks, I should use sage -f more often. So, what became of Jeroen's idea of using the "order only dependency" feature of GNU make? Still a good idea. If Volker considers it a good idea for Sage-6.4, I'll make a patch quickly. -- You received this me

Re: [sage-devel] make behaviour

2014-11-03 Thread Ralf Stephan
Thanks, I should use sage -f more often. So, what became of Jeroen's idea of using the "order only dependency" feature of GNU make? On Mon, Nov 3, 2014 at 10:13 AM, Francois Bissey < francois.bis...@canterbury.ac.nz> wrote: > see https://groups.google.com/forum/#!topic/sage-devel/tyXJC2rtsag > >

Re: [sage-devel] make behaviour

2014-11-03 Thread Francois Bissey
see https://groups.google.com/forum/#!topic/sage-devel/tyXJC2rtsag > On 3/11/2014, at 22:07, Ralf Stephan wrote: > > Hello, > Up to one of the last beta versions, 'make' only built new libraries once. > Now, each time I switch branches to an older one and merge develop, > 'make' will rebuild pyt

Re: [sage-devel] make behaviour

2014-11-03 Thread Francois Bissey
Well, “we” decided that was the only sane behaviour and made it the default. I’ll try to find the thread for you. François > On 3/11/2014, at 22:07, Ralf Stephan wrote: > > Hello, > Up to one of the last beta versions, 'make' only built new libraries once. > Now, each time I switch branches to

[sage-devel] make behaviour

2014-11-03 Thread Ralf Stephan
Hello, Up to one of the last beta versions, 'make' only built new libraries once. Now, each time I switch branches to an older one and merge develop, 'make' will rebuild python-2.7.8, ATLAS and whatnot, taking half an hour just for a branch switch on a 6-core, 8GB desktop machine. As make is still