For reference, see the issue https://github.com/drmeister/clasp/issues/98
To reproduce, do the following:
git clone git://github.com/drmeister/clasp claspcl
cd claspcl
git checkout tags/0.3-test-3
git submodule update --init
The result is:
git submodule update --init
Submodule 'updatedAsdf' (https://github.com/drmeister/asdf.git) registered for path 'src/lisp/kernel/asdf'
Submodule 'src/lisp/kernel/contrib/sicl' (https://github.com/drmeister/SICL.git) registered for path 'src/lisp/kernel/contrib/sicl'
Submodule 'src/mps' (https://github.com/Ravenbrook/mps-temporary) registered for path 'src/mps'
Cloning into 'src/lisp/kernel/asdf'...
remote: Counting objects: 13416, done.
remote: Total 13416 (delta 0), reused 0 (delta 0), pack-reused 13416
Receiving objects: 100% (13416/13416), 5.34 MiB | 838.00 KiB/s, done.
Resolving deltas: 100% (9935/9935), done.
Checking connectivity... done.
Submodule path 'src/lisp/kernel/asdf': checked out 'dcd692a8f9f6a5cb86133123e1c433a9afb4e7e9'
Cloning into 'src/lisp/kernel/contrib/sicl'...
remote: Counting objects: 49356, done.
remote: Total 49356 (delta 0), reused 0 (delta 0), pack-reused 49356
Receiving objects: 100% (49356/49356), 8.49 MiB | 298.00 KiB/s, done.
Resolving deltas: 100% (37020/37020), done.
Checking connectivity... done.
fatal: reference is not a tree: 280c12ab49223c64c6f914944287a7d049cf4dd0
Cloning into 'src/mps'...
remote: Counting objects: 25250, done.
remote: Total 25250 (delta 0), reused 0 (delta 0), pack-reused 25250
Receiving objects: 100% (25250/25250), 86.87 MiB | 831.00 KiB/s, done.
Resolving deltas: 100% (15851/15851), done.
Checking connectivity... done.
Submodule path 'src/mps': checked out '3e6640e7eeb5d086adec18e6227a84da59898e1c'
Unable to checkout '280c12ab49223c64c6f914944287a7d049cf4dd0' in submodule path 'src/lisp/kernel/contrib/sicl'
As you can see, the reproduction recipe does not touch the sicl submodule. So, my best guess is that the clasp Git repository has a spurious reference to 280c12ab49223c64c6f914944287a7d049cf4dd0 somewhere inside it. The folks on #git also seemed to think this was plausible. As you can see, the reproduction recipe does not touch the sicl submodule. So, how best to resolve this problem? What kind of surgery is required for the clasp Git repository?
NOTE: the problem does not show up without the git checkout tags/0.3-test-3 step. So the problem is presumably in that particular branch, here fixnum.