[isabelle-dev] push request (Sublist.thy)

[Makarius]

On Thu, 13 Dec 2012, Dmitriy Traytel wrote:

> I use Mercurial 2.2 and after pushing ed6b40d15d1c the attached error 
> log was generated.

Rewind to the start of the thread ...

According to http://mercurial.selenic.com/wiki/WhatsNew 2.2 introduces 
"parsers: incrementally parse the revlog index in C", see also 
http://selenic.com/hg/annotate/2cdd7e63211b/mercurial/parsers.c which 
looks like the place where these guys started it.

Later there are several amendments, just search "parsers" or "refcount" in 
the WhatsNew.  In 2.2.2 there is even "parsers: fix refcount bug on 
corrupt index", which is probably this changeset 
http://selenic.com/hg/rev/b767382a8675

So this C module might or might not be an explanation for the 
instabilities.


The machines at TUM have mostly Mercurial 2.1.1 or 2.4 now, notably 
macbroy2[0-9].  2.1.1 is before the critical parsers.c, and 2.4 after 
several fixes of it, but it is a "feature release".  After release x.y 
there is usually a x.y.1 and x.y.2 to consolidate.

Ubuntu 12.10 is close to the critical point, but I see already 2.2.2-1 at 
http://packages.ubuntu.com/search?keywords=mercurial&searchon=names&suite=quantal&section=all 
so an update of Ubuntu might help.

My Ubuntu 12.04.1 LTS has the consolidated 2.0.2 before the fatal C 
module.

My current MacPorts snapshot has the consolidated 1.9.2 even before that.


I am not sure if we are already at the point to conclude anything. Somehow 
implicit sliding updates to "latest" versions could be too risky.  The 
curse of being always up-to-date.

Is this the point to go back to hand-compiled Mercurial of some x.y.2 
version?  It was actually quite easy when we started with Mercurial 1.x, 
before it became generally available on the usual OS distributions.


 	Makarius