[bitbake-devel] [PATCH 00/11] bitbake-diffsigs fixes/improvements

[Paul Eggleton]

On Friday, 7 April 2017 8:29:18 PM NZST Paul Eggleton wrote:
> Hi Patrick,
> 
> On Friday, 7 April 2017 7:56:19 PM NZST Patrick Ohly wrote:
> > On Fri, 2017-04-07 at 09:52 +1200, Paul Eggleton wrote:
> > > As part of an effort to add task signature recording and comparison to
> > > buildhistory in OE, I went digging into bitbake-diffsigs and the code
> > > that supports it and discovered that unfortunately it had a number of
> > > bugs - in particular the -t option despite being around for a while and
> > > having numerous band-aids applied in the past still wasn't really
> > > working properly.
> > 
> > I'm not sure whether it is working now, even with your changes. My
> > test, using poky master and your bitbake branch:
> > 
> > 1. bitbake -c cleansstate m4
> > 2. bitbake m4
> > 3. patch m4:
> > diff --git a/meta/recipes-devtools/m4/m4_1.4.18.bb
> > b/meta/recipes-devtools/m4/m4_1.4.18.bb
> > index b12c0adf3ad..0a8e9c10778 100644
> > --- a/meta/recipes-devtools/m4/m4_1.4.18.bb
> > +++ b/meta/recipes-devtools/m4/m4_1.4.18.bb
> > @@ -1,3 +1,7 @@
> > 
> >  require m4-${PV}.inc
> > 
> > +do_install_append () {
> > +    echo hello world
> > +}
> > +
> > 
> >  BBCLASSEXTEND = "nativesdk"
> > 
> > 4. bitbake m4
> > 5. bitbake-diffsigs -t m4 do_build
> > => ERROR: No sigdata files found matching m4 do_build
> > 
> > There is one file for do_build in tmp/stamps/i586-poky-linux/m4 and
> > several for do_install:
> > 
> > $ ls -l tmp/stamps/i586-poky-linux/m4/*do_build*
> > tmp/stamps/i586-poky-linux/m4/*do_install* -rw-r--r-- 1 pohly pohly     0
> > Apr  7 09:53
> > tmp/stamps/i586-poky-linux/m4/1.4.18-r0.do_build.aa7d33bbb2dfef67ef3ca1b5d
> > 3
> > f2cde1 -rw-r--r-- 1 pohly pohly     0 Apr  7 09:53
> > tmp/stamps/i586-poky-linux/m4/1.4.18-r0.do_install.a0954ae23d8da50c4af22f7
> > 1
> > 9255411f -rw-r--r-- 1 pohly pohly 20573 Apr  7 09:52
> > tmp/stamps/i586-poky-linux/m4/1.4.18-r0.do_install.sigdata.464e707d80ec982
> > f
> > b5b0c50a6526f05a -rw-r--r-- 1 pohly pohly 19610 Apr  7 09:53
> > tmp/stamps/i586-poky-linux/m4/1.4.18-r0.do_install.sigdata.a0954ae23d8da50
> > c
> > 4af22f719255411f
> > 
> > 
> > bitbake-diffsigs does not fail for do_install, but does not print
> > anything either (= no output at all from "bitbake-diffsigs -t m4
> > do_install").
> > 
> > Hmm, does it pick the right files to compare?
> > 
> > $ bitbake-diffsigs -d -t m4 do_install
> > DEBUG: Signature file (previous):
> > /fast/build/poky/x86/sstate-cache/a0/sstate:m4:i586-poky-linux:1.4.18:r0:i
> > 5
> > 86:3:a0954ae23d8da50c4af22f719255411f_install.tgz.siginfo DEBUG: Signature
> > file (latest):
> > /fast/build/poky/x86/tmp/stamps/i586-poky-linux/m4/1.4.18-r0.do_install.si
> > g
> > data.a0954ae23d8da50c4af22f719255411f
> > 
> > Am I using the tool incorrectly or are my expectations wrong?
> > 
> > My expectation was that "bitbake-diffsigs -t m4 do_build" would tell me
> > that m4 was rebuilt because some tasks that it depends on changed, and
> > then I would drill down to those tasks, eventually arriving at
> > do_install.
> 
> Hmm, something has gone wrong here because this is the exact scenario it's
> supposed to handle. In my similar tests it did work, so I'm not sure why it
> didn't work in this case - I'll have to give it a try here.

I send and then I immediately realised what might be wrong - you said you're 
using poky master which implies that you haven't also applied the patches for 
lib/oe/sstatesig.py that I sent to the OE-Core list today. If you haven't, try 
doing that, it should fix the problem.

Cheers,
Paul

-- 

Paul Eggleton
Intel Open Source Technology Centre