| Message ID | 20230216105640.147967-1-mikko.rapeli@linaro.org |
|---|---|
| Headers | show |
| Series | switch from :append to += | expand |
On 16 Feb 2023, at 10:56, Mikko Rapeli via lists.yoctoproject.org <mikko.rapeli=linaro.org@lists.yoctoproject.org> wrote: > This same pattern is used with a lot of variables so try to find and fix > all of them. Hopefully there are no regressions due to :append and += > being applied in different order to the result variable. That’s a pretty huge “hopefully”. Did you just blanket change every :append, and did you verify that the parsed datastore is unchanged? Ross
Hi, On Thu, Feb 16, 2023 at 11:15:42AM +0000, Ross Burton wrote: > On 16 Feb 2023, at 10:56, Mikko Rapeli via lists.yoctoproject.org <mikko.rapeli=linaro.org@lists.yoctoproject.org> wrote: > > This same pattern is used with a lot of variables so try to find and fix > > all of them. Hopefully there are no regressions due to :append and += > > being applied in different order to the result variable. > > That’s a pretty huge “hopefully”. Yep :) It's a risk which I'm highlighting. > Did you just blanket change every :append, and did you verify that the parsed datastore is unchanged? I did not verify the data store. I'm in the process of changing our setup from kirkstone to master for a number of boards and this became a blocking issue at recipe parsing time. In our setup we may not be using all these recipes and appends so I would need help validating this. For optee-os recipes, the changes work correctly and we can now update the recipe from our layer. How should I validate this more? A world "bitbake -e" output before and after? Which machines? I did check the details several times and the changes make sense to me. There may be some corner case which I did not spot, for example variable being appended to and then bbclass inherited later which would break with switch to +=. Cheers, -Mikko