Paul Boddie's Free Software-related blog

This kind of thing has happened to me a lot in recent times…

$ git pull
remote: Counting objects: 1367008, done.
remote: Compressing objects: 100% (242709/242709), done.
remote: Total 1367008 (delta 1118135), reused 1330194 (delta 1113455)
Receiving objects: 100% (1367008/1367008), 402.55 MiB | 3.17 MiB/s, done.
Resolving deltas: 100% (1118135/1118135), done.
fatal: missing blob object '715c19c45d9adbf565c28839d6f9d45cdb627b15'
error: git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git did not send all necessary objects

OK, but now what? At least it hasn’t downloaded gigabytes of data and thrown it all away this time. No, just a few hundred megabytes instead.

Looking around on the Internet, I see various guides that are like having your car engine repeatedly stall at traffic lights and being told to crack open the bonnet/hood and start poking at random components, hoping that the thing will jump back to life. After all, isn’t that supposed to be one of the joys of motoring?

Or to translate the usual kind of response about Git when anyone dares to question its usability: “Do you not understand that you are not merely driving a car but are instead interacting with an extensible automotive platform?” And to think that the idea was just to get from A to B conveniently.

I cannot remember a single time when a Mercurial repository failed to update in such a fashion. But I suppose that we will all have to continue to endure the fashionable clamour for projects to move to Git and onto GitHub so that they can become magically popular and suddenly receive a bounty of development attention from the Internet pixies. Because they all love Git, apparently.