Re: [Monotone-devel] bug report: invariant'I(tmp == new_id)' violated

[Nathaniel Smith]

On Sat, Jul 02, 2005 at 07:04:57AM -0700, Otto Sievert wrote:
> Hi,
> 
> On 2 July at ~7am pacific time (US) I was using Monotone on WinXP to pull 
> the colinux source code when I stumbled upon this fatal error. The error is 
> repeatable. This is my first use of monotone, having installed it this 
> morning as well (monotone-0.19-setup.exe). Here is a snippet of the problem:
> 
> monotone: db.fetch("SELECT id FROM 'revisions' WHERE id = 
> 'c41816932e8237b05bc28412a4a5e82877724cee'")
> monotone: database.cc:1442: invariant 'I(tmp == new_id)' violated
> monotone: db.execute("ROLLBACK")
> monotone: fatal: std::logic_error: database.cc:1442: invariant 'I(tmp == 
> new_id)' violated
> monotone:
> monotone: this is almost certainly a bug in monotone.

Monotone is claiming that the server sent it a revision that was
supposed to have one hash, but when it ran SHA1 on that revision, it
came out to have a different hash.  Something very funny going on with
this; if the server's database were itself corrupted, then the server
should have noticed the problem; if the data got corrupted on the
wire, then the wire checksum should have noticed the problem...
people have reported this before with the coLinux repo, but I'm still
at a loss as to what might cause it.

-- Nathaniel

-- 
"Of course, the entire effort is to put oneself
 Outside the ordinary range
 Of what are called statistics."
  -- Stephan Spender