On 2021/7/20 下午4:29, Qu Wenruo wrote:

On 2021/7/20 下午3:57, Eryu Guan wrote:

quoted

On Tue, Jul 20, 2021 at 04:43:17PM +1000, Dave Chinner wrote:
[snip]

quoted

And given that it appears you haven't thought about maintaining a
local repository of hooks, I strongly doubt you've even consider the
impact of future changes to the hook API on existing hook scripts
that devs and test engineers have written over months and years
while debugging test failures.

Darrick pointed out the difference between running in the check vs
test environment, which is something that is very much an API
consideration - we change the test environment arbitrarily and fix
all the tests that change affects at the same time. But if there are
private scripts that depend on the test environment and variables
being stable, then we can't do things like, say, rename the "seqres"
variable across all tests because that will break every custom hook
script out there that writes to $seqres.full...

I was thinking about this as well, if such private hook scripts are
useful to others as well, then I think maybe it's worth to maintain such
scripts in fstests repo, and further changes to the hook API won't break
the scripts

But those hook scripts are really craft by each developer, which can
have vastly different usage.

How could that be maintained inside fstests?

Anyway, if building a stable and complex API just for hooks, then it's
completely against my initial purpose.

I'll just discard this patch then.

Thanks,
Qu

Thanks,
Qu

quoted

Thanks,
Eryu