b4 am support for partial series updates?

b4 am support for partial series updates?

[Dan Williams]

I recently had someone grumble at me about my requirement that they
repost a full series just so "b4 am" could autograb the full set even
though only a handful of patches changed.

As an experiment I tried replying to a patch to update and marked it
v5, but b4 am reports:

# b4 am 162096970332.1865304.10280028741091576940.stgit@dwillia2-desk3.amr.corp.intel.com
...
Writing ./v5_20210514_dan_j_williams_cxl_acpi_introduce_acpi0017_driver_and_cxl_root.mbx
  ERROR: missing [1/8]!
  ERROR: missing [2/8]!
  ERROR: missing [3/8]!
  ERROR: missing [4/8]!
  [PATCH v5 5/8] cxl/acpi: Introduce ACPI0017 driver and cxl_root
  ERROR: missing [6/8]!
  ERROR: missing [7/8]!
  ERROR: missing [8/8]!

...which is helpful if v5 was partially sent, but not if all the other
patches don't require updates. This feels like a "WANTVER+" option
that means "grab latest, but fall back to vX" for missing patches. Has
this come up before?

What format would help b4 detect this situation? By default tooling
would make the prefix be [PATCH], I manually made that [PATCH v5 5/8]
and forced the in-reply-to message-id to the v4 version.

I'm nearing the point where handling this manually is more painful
than figuring out a b4 patch, but please consider this for the b4
feature backlog in the meantime.

Re: b4 am support for partial series updates?

[Konstantin Ryabitsev]

On Fri, May 14, 2021 at 03:50:21PM -0700, Dan Williams wrote:
> I recently had someone grumble at me about my requirement that they
> repost a full series just so "b4 am" could autograb the full set even
> though only a handful of patches changed.
> 
> As an experiment I tried replying to a patch to update and marked it
> v5, but b4 am reports:
> 
> # b4 am 162096970332.1865304.10280028741091576940.stgit@dwillia2-desk3.amr.corp.intel.com
> ...
> Writing ./v5_20210514_dan_j_williams_cxl_acpi_introduce_acpi0017_driver_and_cxl_root.mbx
>   ERROR: missing [1/8]!
>   ERROR: missing [2/8]!
>   ERROR: missing [3/8]!
>   ERROR: missing [4/8]!
>   [PATCH v5 5/8] cxl/acpi: Introduce ACPI0017 driver and cxl_root
>   ERROR: missing [6/8]!
>   ERROR: missing [7/8]!
>   ERROR: missing [8/8]!
> 
> ...which is helpful if v5 was partially sent, but not if all the other
> patches don't require updates. This feels like a "WANTVER+" option
> that means "grab latest, but fall back to vX" for missing patches. Has
> this come up before?
> 
> What format would help b4 detect this situation? By default tooling
> would make the prefix be [PATCH], I manually made that [PATCH v5 5/8]
> and forced the in-reply-to message-id to the v4 version.
> 
> I'm nearing the point where handling this manually is more painful
> than figuring out a b4 patch, but please consider this for the b4
> feature backlog in the meantime.

I was generally against handling such situations in the past because
they introduced ambiguity. E.g. consider this two-patch series:

[PATCH v2 1/2] First patch
\- [PATCH v3 1/2] First patch
[PATCH v2 2/2] Second patch

It's unclear here what happened:

- is it a single-patch reroll as you describe above?
- or is it a full reroll and [PATCH v3 2/2] simply got lost en route?

However, we are probably safe to treat it as a partial patch reroll if
the following conditions are true:

1. The series has a cover letter
2. Individual patch rerolls are sent as follow-ups to exact same patch
X/N (e.g. [PATCH v3 3/5] is sent as a follow-up to [PATCH v2 3/5].

Let me brainstorm about this on Monday and I'll share what I came up
with once I have something to show.

Best regards,
-K

Re: b4 am support for partial series updates?

[Konstantin Ryabitsev]

On Fri, May 14, 2021 at 03:50:21PM -0700, Dan Williams wrote:
> I recently had someone grumble at me about my requirement that they
> repost a full series just so "b4 am" could autograb the full set even
> though only a handful of patches changed.
> 
> As an experiment I tried replying to a patch to update and marked it
> v5, but b4 am reports:
> 
> # b4 am 162096970332.1865304.10280028741091576940.stgit@dwillia2-desk3.amr.corp.intel.com
> ...
> Writing ./v5_20210514_dan_j_williams_cxl_acpi_introduce_acpi0017_driver_and_cxl_root.mbx
>   ERROR: missing [1/8]!
>   ERROR: missing [2/8]!
>   ERROR: missing [3/8]!
>   ERROR: missing [4/8]!
>   [PATCH v5 5/8] cxl/acpi: Introduce ACPI0017 driver and cxl_root
>   ERROR: missing [6/8]!
>   ERROR: missing [7/8]!
>   ERROR: missing [8/8]!
> 

Okay, the latest master will handle this situation:

    Looking up https://lore.kernel.org/r/162096970332.1865304.10280028741091576940.stgit%40dwillia2-desk3.amr.corp.intel.com
    Grabbing thread from lore.kernel.org/linux-acpi
    Analyzing 13 messages in the thread
    Partial reroll detected, reconstituting from v4
    Will use the latest revision: v5
    You can pick other revisions using the -vN flag
    ---
    Writing /tmp/v5_20210513_dan_j_williams_cxl_port_enumeration_and_plans_for_v5_14.mbx
    Checking attestation on all messages, may take a moment...
      [PATCH v4->v5 1/8] cxl/mem: Move some definitions to mem.h
      [PATCH v4->v5 2/8] cxl/mem: Introduce 'struct cxl_regs' for "composable" CXL devices
      [PATCH v4->v5 3/8] cxl/core: Rename bus.c to core.c
      [PATCH v4->v5 4/8] cxl/core: Refactor CXL register lookup for bridge reuse
      [PATCH     v5 5/8] cxl/acpi: Introduce ACPI0017 driver and cxl_root
      [PATCH v4->v5 6/8] cxl/Kconfig: Default drivers to CONFIG_CXL_BUS
      [PATCH     v5 7/8] cxl/port: Introduce cxl_port objects
      [PATCH v4->v5 8/8] cxl/acpi: Add module parameters to stand in for ACPI tables
    ---
    Total patches: 8
    ---
    WARNING: v5 is a partial reroll from previous revisions
             Please carefully review the resulting series to ensure correctness
             Pass --no-partial-reroll to disable

Best regards,
-K

Re: b4 am support for partial series updates?

[Dan Williams]

On Mon, May 17, 2021 at 9:13 AM Konstantin Ryabitsev
<konstantin@linuxfoundation.org> wrote:
>
> On Fri, May 14, 2021 at 03:50:21PM -0700, Dan Williams wrote:
> > I recently had someone grumble at me about my requirement that they
> > repost a full series just so "b4 am" could autograb the full set even
> > though only a handful of patches changed.
> >
> > As an experiment I tried replying to a patch to update and marked it
> > v5, but b4 am reports:
> >
> > # b4 am 162096970332.1865304.10280028741091576940.stgit@dwillia2-desk3.amr.corp.intel.com
> > ...
> > Writing ./v5_20210514_dan_j_williams_cxl_acpi_introduce_acpi0017_driver_and_cxl_root.mbx
> >   ERROR: missing [1/8]!
> >   ERROR: missing [2/8]!
> >   ERROR: missing [3/8]!
> >   ERROR: missing [4/8]!
> >   [PATCH v5 5/8] cxl/acpi: Introduce ACPI0017 driver and cxl_root
> >   ERROR: missing [6/8]!
> >   ERROR: missing [7/8]!
> >   ERROR: missing [8/8]!
> >
>
> Okay, the latest master will handle this situation:
>
>     Looking up https://lore.kernel.org/r/162096970332.1865304.10280028741091576940.stgit%40dwillia2-desk3.amr.corp.intel.com
>     Grabbing thread from lore.kernel.org/linux-acpi
>     Analyzing 13 messages in the thread
>     Partial reroll detected, reconstituting from v4
>     Will use the latest revision: v5
>     You can pick other revisions using the -vN flag
>     ---
>     Writing /tmp/v5_20210513_dan_j_williams_cxl_port_enumeration_and_plans_for_v5_14.mbx
>     Checking attestation on all messages, may take a moment...
>       [PATCH v4->v5 1/8] cxl/mem: Move some definitions to mem.h
>       [PATCH v4->v5 2/8] cxl/mem: Introduce 'struct cxl_regs' for "composable" CXL devices
>       [PATCH v4->v5 3/8] cxl/core: Rename bus.c to core.c
>       [PATCH v4->v5 4/8] cxl/core: Refactor CXL register lookup for bridge reuse
>       [PATCH     v5 5/8] cxl/acpi: Introduce ACPI0017 driver and cxl_root
>       [PATCH v4->v5 6/8] cxl/Kconfig: Default drivers to CONFIG_CXL_BUS
>       [PATCH     v5 7/8] cxl/port: Introduce cxl_port objects
>       [PATCH v4->v5 8/8] cxl/acpi: Add module parameters to stand in for ACPI tables
>     ---
>     Total patches: 8
>     ---
>     WARNING: v5 is a partial reroll from previous revisions
>              Please carefully review the resulting series to ensure correctness
>              Pass --no-partial-reroll to disable

Thank you, much appreciated!