Is the build system SCM sensitive?
I had a complete build (details probably don't matter) using Poky master. I'm working on a fix to one of the core recipes, so I made a local branch:
% git checkout -b fix-python-pygtk master
I made a single line change in the recipe (removing a line from the do_install step). When I then rebuilt the recipe, to my surprise, there were more than 450 tasks (73 unique recipes) executed :-(
How does this make any sense? unless the build system cares that I changed the branch?
n.b. I'm happy to provide more details if necessary. Also, I did this twice in two different build trees and saw the same strangeness.
Signature · 4 lines
------------------------------------------------------------ Gary Thomas | Consulting for the MLB Associates | Embedded world ------------------------------------------------------------
Quoted 9 lines · Gary Thomas
I had a complete build (details probably don't matter) using Poky master. I'm working on a fix to one of the core recipes, so I made a local branch:
% git checkout -b fix-python-pygtk master
I made a single line change in the recipe (removing a line from thedo_installstep). When I then rebuilt the recipe, to my surprise, there were more than 450 tasks (73 unique recipes) executed :-(
Quoted 7 lines
How does this make any sense? unless the build system cares that I changed the branch?
n.b. I'm happy to provide more details if necessary. Also, I did this twice in two different build trees and saw the same strangeness.
I think so additional info might be needed here, I am guessing that you are changing pygtk and somehow that is changing some dependency, you might try using bitbake-diffsigs between the 2 versions.
Sau!
Quoted 13 lines · Saul Wold
Quoted 12 lines · Gary Thomas
I had a complete build (details probably don't matter) using Poky master. I'm working on a fix to one of the core recipes, so I made a local branch:
% git checkout -b fix-python-pygtk master
I made a single line change in the recipe (removing a line from thedo_installstep). When I then rebuilt the recipe, to my surprise, there were more than 450 tasks (73 unique recipes) executed :-(How does this make any sense? unless the build system cares that I changed the branch?
Quoted 6 lines
Quoted 4 lines
n.b. I'm happy to provide more details if necessary. Also, I did this twice in two different build trees and saw the same strangeness.
I think so additional info might be needed here, I am guessing that you are changing pygtk and somehow that is changing some dependency, you might try using bitbake-diffsigs between the 2 versions.
I think I know what triggered this - a recent change in bitbake.conf redefined BUILD_CPP (and others). I had merged this change some six hours before my strange build and had been building many other recipes (also python with a similar set of dependencies) a number of times before I touched the python-pygtk recipe. Some dependency in that recipe set off the rebuild storm that my other recipes had not.
So the large number of recipe rebuilds was [probably] warranted, but very unexpected given what I had been doing just prior.
Sorry for the noise. Next time I'll try and remember bitbake-diffsigs and figure things out on my own.
Signature · 4 lines
------------------------------------------------------------ Gary Thomas | Consulting for the MLB Associates | Embedded world ------------------------------------------------------------
Quoted 7 lines · Gary Thomas
I think I know what triggered this - a recent change in bitbake.conf redefined
BUILD_CPP(and others). I had merged this change some six hours before my strange build and had been building many other recipes (also python with a similar set of dependencies) a number of times before I touched the python-pygtk recipe. Some dependency in that recipe set off the rebuild storm that my other recipes had not.
Ah yes, welcome back to master. There's been stuff building up for a while. :)
This is where I point out that master between a release and M1 is known to be "volatile". Whilst we obviously endeavour for it to be always buildable, deep and fundamental changes will be going in sooner rather than later.
(ie this might be time to dust of my libexecdir-changing series)
Ross
Quoted 11 lines · Ross
On 27 October 2014 19:40, Gary Thomas <[email protected] <mailto:[email protected]>> wrote:
I think I know what triggered this - a recent change in bitbake.conf redefined
BUILD_CPP(and others). I had merged this change some six hours before my strange build and had been building many other recipes (also python with a similar set of dependencies) a number of times before I touched the python-pygtk recipe. Some dependency in that recipe set off the rebuild storm that my other recipes had not.
Quoted 4 lines
Ah yes, welcome back to master. There's been stuff building up for a while. :)
This is where I point out that master between a release and M1 is known to be "volatile". Whilst we obviously endeavour for it to be always buildable, deep and fundamental changes will be going in sooner rather than later.
This doesn't bother me - I expect it. What took me by surprise was when the behaviour radically changed from what I had seen all day. Totally explainable, but unexpected nonetheless.
Quoted 2 lines
(ie this might be time to dust of my libexecdir-changing series)
Go for it :-)
Signature · 4 lines
------------------------------------------------------------ Gary Thomas | Consulting for the MLB Associates | Embedded world ------------------------------------------------------------