On Tue, Jan 6, 2026 at 2:15 AM Rob Landley <[email protected]> wrote: > > On 1/5/26 15:11, enh wrote: > > does `git pull` not fix it? > > I did a new --force push by hand and it seems to have worked. Probably > just a transient connection failure I didn't notice at the time. (The > downside of dialing out multiple times in the same transaction, it says > on the command line if there's an error but for all I know I closed the > tab early...) > > I was avoiding just --force pushing again because I didn't want to screw > up things on your end... but you'll just wind up with an extra merge > commit and a stranded "dummy" file in your repo, shouldn't be a problem. > You _already_ merge each pull.
yeah, it's actually sometimes mildly useful to have the "bogus" merge commits in the android history because it makes it clear what points of upstream history were ever actually in use android. (and github's bizarre workflow is so prone to random merge cruft anyway that i've always just checked out a new clone for every patch in projects i don't own.) > (The theory behind the script was "it's literally there for 5 seconds, > the likelihood of anybody cloning at the wrong time is miniscule. The 5 > second wait is because if you --force push the new commit away too fast > microsoft github's plumbing won't acknowledge the commit was "to the > main branch", which is apparently necessary to close the issue. I very > much do not WANT to have to know or understand any of this, but here we > are...) > > > (and you can still just mail me any bugs you want closed :-) ) > > I may start doing that. :) > > Thanks, > > Rob > > P.S. I've never pulled from an external source in the history of the > toybox repo, but that's just because I don't want to sit down and think > through the security implications of somebody engineering an sha1sum > collision (https://valerieaurora.org/hash.html doesn't take state actors > into account, let alone palantir/blackwater/wagner type nonsense and > whatever faang keeps to themselves) or doing something clever with an > orphan commit. (I _could_ do commit signing with PGP keys and such, and > people out there who HAVE thought through this think they need to, but > that katamari only grows.) There are already a couple merge commits in > the toybox history (one was my screw up and one was me trying a new > workflow with a friend and deciding I didn't like it), adding a third > wouldn't be a deal breaker but I want microsoft github to remain an > _output_ not an input. Having microsoft github-only commits in the tree > gives them too much importance, hence the script having a --force push > in the first place. (And yes I'm aware git log .github gives 10 commits, > but "commit to close microsoft github issue" boils a new frog.) _______________________________________________ Toybox mailing list [email protected] http://lists.landley.net/listinfo.cgi/toybox-landley.net
