x/build/cmd/gopherbot: -2 any CLs with large binary blobs unless declared in commit message

[josharian]

See CL 9560 for context.

There will be cases in which it is legit to add a binary file. In such cases, we could do a force commit. Other suggestions welcomed.

Edit by @dmitshur: This also happened in CL 149604, see #28899.

[bradfitz]

Instead of a force commit, we can say that adding a large binary files requires an explicit magic phrase in the commit message, as acknowledgment of intent.

[nightlyone]

A simple thing would be, to list the to be added binary files in the commit message, to make it visible and communicate intent.

[mwhudson]

I don't know if git-codereview is the best place to enforce this, as its use is not mandatory (I don't use it any more, for example), Gerrit would be better. Codereview would be better than nothing, mind.

[josharian]

Slightly off-topic, but why not, @mwhudson? What do you do instead?

If this were added, it would be a pre-commit hook, so if you use the hooks, it'd be picked up.

I agree that we should enforce some simple size rules on the gerrit side, but doing it here too lets us put in more subtle and more restrictive detection logic, which the user can intentionally overrule. I like @nightlyone's suggestion for that. Overruling a gerrit check is way beyond a reviewer sanity-check.

[mwhudson]

On 4 May 2015 at 05:26, Josh Bleecher Snyder notifications@github.com
wrote:

Slightly off-topic, but why not, @mwhudson https://github.com/mwhudson?
What do you do instead?

I think it's because I'm a medium skill git/gerrit user. If I was an
expert, I'd understand what codereview is doing in git terms, if I was a
novice I'd probably just use codereview, but at my level, I find it
confusing -- especially 'git mail' and especially when I was working with
that stack of 15 changes for shared library support. Instead, well, I just
use git (git mail becomes "git push origin HEAD:refs/for/master" and
somehow I find that easier to understand).

If this were added, it would be a pre-commit hook, so if you use the
hooks, it'd be picked up.

Ah yes I still use the hooks (although given I have "(add-hook
'before-save-hook 'gofmt-before-save)" in .emacs, all it does over the
standard gerrit hook is tell me I can't commit on master from time to time).

[bradfitz]

@josharian,

If this were added, it would be a pre-commit hook, so if you use the hooks, it'd be picked up.

We should do this server-side (in gopherbot-ish things) because we accept GitHub PRs too now. We should do #18548 instead.

/cc @andybons

[josharian]

SGTM. #18548 is pretty busy. Let’s leave this open, so it doesn’t get lost in the noise of that issue. M

[bradfitz]

I agree we should leave this open. This is a feature that should be implemented as part of that system.