Skip to contentSkip to conversation

kernel-yocto.bbclass ignoring SRCREV?

DISCUSSION Wolfgang Denk 2 participants 3 messages May 21, 2014

lore
Share & tools
lore
#0 lore raw

Hi,

is my understanding of kernel-yocto.bbclass correct that it effectively completely ignores any specific git commit ID that was gien in SRCREV, but instead always checks out and uses the HEAD of the respective branch?

Or am I missing something here? My expectation was that we specify a git commit ID (as the only reliable, truly unique identifier for a specific source code version) in SRCREV, but instead it appears that this information is effectively being ignored?

I can understand that there are situations where automatically using the top commit of a branch (say, the latest stable version of some package) may be useful, but should there not be some different way to do that?

I mean, how can we make sure to ever be able to reproduce the very same build if the git commit ID specified in SRCREV will not be used for the checkout?

Thanks in advance.

Best regards,

Wolfgang Denk

Signature · 6 lines
DENX Software Engineering GmbH,     MD: Wolfgang Denk & Detlev Zundel
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: [email protected]
"More software projects have gone awry for lack of calendar time than
for all other causes combined."
                         - Fred Brooks, Jr., _The Mythical Man Month_
#1 lore raw
Quoted 11 lines · Wolfgang Denk

Hi,

is my understanding of kernel-yocto.bbclass correct that it effectively completely ignores any specific git commit ID that was gien in SRCREV, but instead always checks out and uses the HEAD of the respective branch?

Or am I missing something here? My expectation was that we specify a git commit ID (as the only reliable, truly unique identifier for a specific source code version) in SRCREV, but instead it appears that this information is effectively being ignored?

kernel-yocto does respect SRCREV. It uses the routine do_validate_branches to reset to a specific SRCREV as a baseline, it then starts processing and applying BSP descriptions on top of that point.

We test SRCREV specific builds on a ongoing basis, so they do work, and if a different behaviour is being seen .. that's a bug.

Cheers,

Bruce

Quoted 16 lines

I can understand that there are situations where automatically using the top commit of a branch (say, the latest stable version of some package) may be useful, but should there not be some different way to do that?

I mean, how can we make sure to ever be able to reproduce the very same build if the git commit ID specified in SRCREV will not be used for the checkout?

Thanks in advance.

Best regards,

Wolfgang Denk

#2 lore raw
Quoted 8 lines · Bruce Ashfield
Quoted 7 lines · Wolfgang Denk

Hi,

is my understanding of kernel-yocto.bbclass correct that it effectively completely ignores any specific git commit ID that was gien in SRCREV, but instead always checks out and uses the HEAD of the respective branch?

Quoted 8 lines
Quoted 4 lines

Or am I missing something here? My expectation was that we specify a git commit ID (as the only reliable, truly unique identifier for a specific source code version) in SRCREV, but instead it appears that this information is effectively being ignored?

kernel-yocto does respect SRCREV. It uses the routine do_validate_branches to reset to a specific SRCREV as a baseline, it then starts processing and applying BSP descriptions on top of that point.

And I forgot to point out that do_validate_branches() has been nearly completely removed/rewritten as part of the upcoming yocto 1.7 development cycle. So what I mention above is more clear in the new versions.

Bruce

Quoted 25 lines

We test SRCREV specific builds on a ongoing basis, so they do work, and if a different behaviour is being seen .. that's a bug.

Cheers,

Bruce

Quoted 16 lines

I can understand that there are situations where automatically using the top commit of a branch (say, the latest stable version of some package) may be useful, but should there not be some different way to do that?

I mean, how can we make sure to ever be able to reproduce the very same build if the git commit ID specified in SRCREV will not be used for the checkout?

Thanks in advance.

Best regards,

Wolfgang Denk