[PATCH 0/4] Fix a race condition when reading loose refs

This patch series is inspired by the first patch [1] in a series by Peff and solves the race condition that he discovered plus a second, similar race. The general problem is that loose ref files are read in two steps: lstat() followed by readlink() or open(). If another process changes the file between our two steps, then the old code gets confused and can give incorrect results. The new code detects if an error reported by readlink()/open() is inconsistent with the results of the previous lstat(), and if so, tries anew.

The first three commits are just preparatory. All four commits could easily be squashed together if it is preferred (especially commits 2 through 4), but I did it this way to separate the code-churning-but-trivial changes from the functional changes to make review easier.

[1] http://article.gmane.org/gmane.comp.version-control.git/223525

Michael Haggerty (4):
  resolve_ref_unsafe(): extract function handle_missing_loose_ref()
  resolve_ref_unsafe(): handle the case of an SHA-1 within loop
  resolve_ref_unsafe(): nest reference-reading code in an infinite loop
  resolve_ref_unsafe(): close race condition reading loose refs

 refs.c | 187 ++++++++++++++++++++++++++++++++++++++++-------------------------
 1 file changed, 114 insertions(+), 73 deletions(-)