Hi Samuel, On Sat, Dec 25, 2021 at 08:17:01AM -0300, Samuel Henrique wrote: > I'm not familiar with how frequent people are doing architecture > bootstraps on Debian, so would you like me to revert that specific > change?
jenkins.debian.net bootstraps Debian around 10 times a day. I asked you to (temporarily) revert the change via #1002597. > I can surely do it with no problem, I'm asking because you already > have a debdiff and now you might say it's ok to wait for an NMU > (again, I don't know how grave it would be to block bootstraping in > the meanwhile). The libssh upload certainly is insufficient for fixing architecture bootstrap. It is preparing libssh for being added to the bootstrap, but that addition is a separate step. It hasn't been tested thus far, but I don't expect much problems there. The issue with keeping architecture bootstrap in a broken state is that we become blind to other breakage. With some recurrence, architecture has been broken while it was broken by something else. Figuring out the cause of the breakage is much easier when it was in a working state before. So the main downside of keeping it broken is me spending way more time when something else also breaks bootstrap. > I also don't wanna give you more work than you already have, so if you > don't reply soon (+/-18h), I will go ahead with the revert (since this > is the safest approach). I'm unsure here. Martin already commited my libssh patch to git (thank you!) and intends to upload soon. It seems to me that the long term goal still is to switch to libssh for curl. If both is true, I prefer adding libssh to the bootstrap sooner rather than later, because we will then see libssh regressions wrt bootstrap (if any) via jenkins. Helmut
