| Summary: | [Usability] Way to determine what is about to happen (similar to a dry-run) | ||
|---|---|---|---|
| Product: | [Build System, Metadata & Runtime] OE-Core | Reporter: | Mark Hatle <mark.hatle> |
| Component: | devtools / tool chain | Assignee: | Robert Yang <liezhi.yang> |
| Status: | VERIFIED FIXED | QA Contact: | |
| Severity: | enhancement | ||
| Priority: | Medium | CC: | corneliux.stoicescu, jessica.zhang, liezhi.yang, meta.mr.watcher, meta.watcher, poky.bs.watcher, poky.watcher, richard.purdie, sgw, song.liu, zhenfeng.zhao |
| Version: | 1.4 | ||
| Target Milestone: | 1.4 M5 | ||
| Hardware: | x86 | ||
| OS: | Multiple | ||
| Whiteboard: | |||
| OS type for building Yocto: | --- | Type of Regression: | --- |
| Verified: | Documentation change: | --- | |
|
Description
Mark Hatle
2011-10-05 12:12:12 UTC
Is -n what your looking for? -n does not appear to be what I am referring to. -n will simply do a dry run and tell you what is going to happen, but it won't tell you why. What I'd like to see is an argument that will say: do_fetch, run because version/pr changed do_patch, run because SRC_URI references changed do_compile, run because chksum of do_compile changed That way when you do a git pull (or bring in a layer, or...) you can run a command and figure out what is going to happen, and why it's going to happen.. this will let you see if you really want to do it -- or if you are trying to figure out why each time you build some task keeps executing, when you don't think it should. Hi Richard, I re-routed this feature to the "Product: Bitbake", would you please help me to take a look at the design? (The Queues in bitbake/lib/bb/runqueue.py are so complicated). Here is the design doc: 1) This is similar to "bitbake -n", I think that we don't have to add a new option, just make an improvement to the "bitbake -n" is OK, the "bitbake -n" goes through the motions, and show a summary in the end currently (e.g.: bitbake -n core-image-sato): NOTE: Running noexec task 5459 of 5459 (ID: 10, /buildarea2/lyang1/poky/meta/recipes-sato/images/core-image-sato.bb, do_build) NOTE: Tasks Summary: Attempted 5459 tasks of which 5457 didn't need to be rerun and all succeeded. We can the print why it needs to re-run, e.g.: NOTE: (/buildarea2/lyang1/poky/meta/recipes-sato/images/core-image-sato.bb, do_build) needs re-run xxx 2) Edit check_stamp_task or add a similar function to return the reason why need re-run, and print the it when dry-run (-n) 3) The message will be printed by logger.info in lib/bb/runqueue.py I'm going to aim a bit higher here and propose a bitbake-whatchanged script. This would basically: a) run something like STAMPS=/some/tempdir bitbake xxx -S This would generate the siginfo files for what bitbake xxx would build (without actually building it). b) Compare the output in /some/tempdir with that in STAMPS and the hashes availble in the sstate cache. The default output would probably be something like a count of the recipes that would be built, a count of those installed from sstate and count of things that are unchanged. Extra parameters to bitbake-whatchanged would allow you to dive into the details of why any given thing was getting rebuilt, down to providing the output of bitbake-diffisgs against the appropriate files. Re-route it to oe-core since it will go into oe-core/scripts/ Move it to 1.4 since the name of the nativesdk changed. Merged: http://git.yoctoproject.org/cgit/cgit.cgi/poky/commit/?id=1e6d0da8d5456d04c2354d37900a59d292bcd898 http://git.yoctoproject.org/cgit/cgit.cgi/poky/commit/?id=4d5ef2af504635c60ffd2d971191bc111035c3a2 Verified in master:9ffee9add2d0380f4025e7a678a166f25a7d659e Test steps: 1) "bitbake man" 2) add "x11vnc" to "DEPENDS" in "meta/recipes-extended/man/man_1.6g.bb" 3) "bitbake-whatchanged -v man" Result: Output describes what needs to be done/redone over the previous man build. test case created: https://bugzilla.yoctoproject.org/tr_show_case.cgi?case_id=754 |