Re: git: 1dec3639fd0c - main - sysutils/u-boot: Update to 2021.07

From: Mark Millard via freebsd-arm <freebsd-arm_at_freebsd.org>
Date: Sun, 11 Jul 2021 00:29:57 UTC
On 2021-Jul-10, at 16:52, Mark Millard <marklmi@yahoo.com> wrote:

> On 2021-Jul-10, at 11:37, Herbert J. Skuhra <herbert at gojira.at> wrote:
> 
>> On Sat, 10 Jul 2021 15:42:57 +0200, "Herbert J. Skuhra" wrote:
>>> 
>>> My RPi2 boots with U-Boot 2021.01. So does it with:
>>> 
>>> U-Boot 2021.04-rc1 OK efipart_readwrite: rw=1, blk=62518264 size=8 status=2
>>> U-Boot 2021.04-rc2 OK efipart_readwrite: rw=1, blk=62518264 size=8 status=2
>>> U-Boot 2021.04-rc3 OK
>>> U-Boot 2021.04-rc4 OK
>>> U-Boot 2021.04-rc5 OK
>>> U-Boot 2021.04 OK
>>> U-Boot 2021.07 OK
>>> 
>>> But starting with 2021.04-rc4 it only works if you do a soft/warm
>>> boot.
>> 
>> The problem starts with commit "78af81c345430a9088235f48d302922955d2499f".
>> 
>> https://github.com/u-boot/u-boot/archive/63af92e837f3d7c21ab5fc4a96ffcbf202efaf90.zip
>> ==> OK
> 
> https://github.com/u-boot/u-boot/commit/63af92e837f3d7c21ab5fc4a96ffcbf202efaf90
> shows it as:
> 
> master
> v2021.07 ... v2021.07-rc1
> committed on Mar 2
> 
> but clicking the ... shows the other tags that include the commit:
> 
> • v2021.07-rc5
> 
> • v2021.07-rc4
> 
> • v2021.07-rc3
> 
> • v2021.07-rc2
> 
> • v2021.07-rc1
> 
> In other words, not added to master until fairly
> recently (v2021.07-rc1).
> 
> (The commit to some branch was on Mar 2, not the
> merge to main. I do not know how to find the
> original branch.)
> 
>> https://github.com/u-boot/u-boot/archive/78af81c345430a9088235f48d302922955d2499f.zip
>> ==> NOK
> 
> https://github.com/u-boot/u-boot/commit/78af81c345430a9088235f48d302922955d2499f
> shows it as:
> 
> master
> v2021.07 ... v2021.07-rc4
> committed on Mar 2
> 
> (The commit to some branch was on Mar 2, not the
> merge to main. The Mar 2 commit was to some different
> branch than 63af92e837f3 was originally committed to.
> I do not know how to find the original branches.)
> 
> but clicking the ... shows the other tags that include the commit:
> 
> • v2021.07-rc5
> 
> • v2021.07-rc4
> 
> • v2021.07-rc3
> 
> • v2021.07-rc2
> 
> • v2021.07-rc1
> 
> • v2021.04
> 
> • v2021.04-rc5
> 
> • v2021.04-rc4
> 
> In other words, added to master earlier than
> 63af92e837f3 was added to master --by a fair amount.
> 
> 
> 
> https://github.com/u-boot/u-boot/commits/63af92e837f3d7c21ab5fc4a96ffcbf202efaf90
> 
> does not show 78af81c34543 in the history.
> 
> https://github.com/u-boot/u-boot/commit/78af81c345430a9088235f48d302922955d2499f
> 
> does not show 63af92e837f3 in the history.
> 
> (The 2 displays are not based on the ordering of commits
> to master but based on the original branches from what
> I can tell.)
> 
> 
> 
> So, relative to master, there are a lot of commits
> between the two --and the "good" one comes after
> the "bad" one on master.
> 
> I'm guessing that the the bisect should be based on
> the order of inclusion into master instead of spanning
> multiple parallel/independent branches from before the
> merge to master.
> 
> (Such complications are part of why my time allocation
> has not gone into this bisect so far.)
> 

Some more context from looking around (I'm still trying
to get use to git ways of finding things):

https://github.com/u-boot/u-boot/commits/v2021.04-rc4?after=4103e13534141c31e4e9bf40848ab3a61dabce81+139&branch=v2021.04-rc4

indicates that 78af81c34543 was the 7th commit into
v2021.04-rc4 ( i.e., after v202104-rc3 ).

So if v2021.04-rc3 worked completely normally for the issue
at hand, then . . .

v2021.04-rc3 in master was followed by (in sequence):

60a376b: configs: RPi2: Disable EFI Grub workaround
76cffd5: drivers: mmc: iproc_sdhci: enable HS200 mode
0267539: MAINTAINERS: Update info for Raspberry Pi
3ac0523: rng: iproc_rng200: Enable support for RPi4 armv7
fae165b: configs: rpi4_32: Enable iProc RNG200
20ecfbe: configs: Resync with savedefconfig
78af81c: Merge tag 'rpi-next-2021.04.2' of https://source.denx.de/u-boot/custodians/u-boot-raspberrypi

That would appear to be the bisect range relative to
commits to master. Beyond that it would appear finding
the original branch history that the then0identified
commit is from might lead to more bisecting there if
the commit to master is large/complicated.

===
Mark Millard
marklmi at yahoo.com
( dsl-only.net went
away in early 2018-Mar)