u-boot: check the produced platform_install.sh (created from functions like write_uboot_platform) for shellcheck errors

[rpardini]

u-boot: check the produced platform_install.sh (created from functions like write_uboot_platform) for shellcheck errors

• u-boot: check the produced platform_install.sh (created from functions like write_uboot_platform) for shellcheck errors
  • we've some smelly stuff in write_uboot_platform for some families that we'd rather catch early
  • implement small syntax fixes in setup_write_uboot_platform

[igorpecovnik]

why is using local not good?

[rpardini]

why is using local not good?

unclear what you're asking here? what local, what good?

[EvilOlaf]

I assume the question was about why local has been removed in a few places like https://github.com/armbian/build/pull/7509/files#diff-97ee188e57fa0d1ae8bc10a822bcb1bf7ec9fd3e66697c3a6c385f045513dce3R309

[igorpecovnik]

Yes, local variable in function vs global. Probably makes no diff. just asking if there is a reasoning for this.

[rpardini]

I assume the question was about why local has been removed in a few places like #7509 (files)

local variable in function vs global

That's not about local vs global. That's about the logic and specially the error control flow.

A statement with local x=1 does two things: 1) set the value of that variable and 2) declare a local variable. Those two things can fail, but in that specific order.

The return value ("success") of the local x=1 statement will be always true, because no matter if setting the value worked or not, declaring the variable as local always works. (Well not really, but for now lets say so).

Now for such simple (local x=1) statement that makes little difference. Let's take the case @EvilOlaf pointed out, despite that being horrible code:

local part=$(findfs PARTUUID=$tmp 2> /dev/null) # always works, even if findfs failed

So here, if findfs fails, we want to fail. But also here, local part will alway succeed, thus swallowing the error and driving people insane.

Thus one can separate the declaration of the variable scope in a statement, and the assignment in another, so the errors are not hidden.

local part # always works
part=$(findfs PARTUUID=$tmp 2> /dev/null) # will fail if findfs fails

I just realized I wrote a bad version of https://www.shellcheck.net/wiki/SC2155 so go read that instead.

---

Finally, why sometimes local and sometimes declare? That's a bit of bashism, I think, but declare also allows for extra modifiers (eg -r, -g, -A, -a and others).

[The-going]

Has a full check of this code been done?
I mean, to assemble an image and run it on the device.

[rpardini]

on the device.

Which device? I've tested the ones I have (rockchip64 and meson64).