{"thread":{"id":"31974","subject":"Empty name in fast-import stream.","startedAt":"2012-10-30T00:16:30Z","lastAt":"2012-10-30T00:16:30Z","messageCount":1,"participants":["Alexander Usov"],"isPatch":false,"patchVersion":null,"patchTotal":null},"messages":[{"id":"202159","messageId":"CAH_EFyZc3gjOVDKKeVAFuU_KMtf9utr=jEaiRfjujtkTZAGoAA@mail.gmail.com","threadId":"31974","inReplyTo":null,"subject":"Empty name in fast-import stream.","fromName":"Alexander Usov","fromEmail":"a.s.usov@gmail.com","sentAt":"2012-10-30T00:16:30Z","receivedAt":"2012-10-30T00:16:30Z","isPatch":false,"sender":{"key":"a.s.usov@gmail.com","avatar":null},"body":"Hi,\n\nI'm working now on converting large repo to git, and stumbled upon a\ncase where fast-import from old git creates incorrect commit objects.\nThe case is following: when exporting the commit where the only thing\nI know about committer is email, fast-import spec dictates following\nformat:\n\ncommitter <foo@bar.com> 123456789 +100\n\nwith just 1 space between committer and opening brace. Apparently\ngit-fast-import in 1.7.1 would produce a commit object with just 1\nspace in there too. And git-fsck in newer versions of git would\ncomplain about this commit object.\n\nThe workaround I found is to emit 2 spaces between committer and\nopening bracket. Such stream is accepted by both 1.7.1 and 1.8 and\nthey both produce correct commit objects.\n\nI'm wondering if it is worth changing git-fast-import manual page to\nhave following rule for committer (and author):\n\n('author' SP <name>? SP LT <email> GT SP <when> LF)?\n'committer' SP <name>? SP LT <email> GT SP <when> LF\n\nThis format results in the fast-import streams which are correctly\nhandled by both old and newer git versions.\n\n-- \nBest regards,\n  Alexander.\n"}]}