Closed nwf closed 8 years ago
Thanks, both look good to me.
merged. thanks!
btw, what system/machine do you have that it causes crashes? i'd like to add a test for that.
I am running a Debian hybrid of jessie/testing/experimental on amd64, using ZFS on Linux to provide storage (but I do not think that ZFS, while exotic-ish, has factored into any of the bugs I have encountered). Currently a 4.7 kernel. Note that I have built from HEAD and am not using the Debian package which, for unknown reasons, failed to be compiled with UNIONFS_HAVE_AT and so doesn't pick up the availability of utimensat() and so attempts to follow symlinks for utimes() calls, which, well, doesn't work out well in general (e.g. when symlinks are to root-owned targets like /tmp).
All my testing has been "build, compose a unionfs, and try to build LEDE/OpenWRT inside the union mount". The findbranch bug was tickled in particular by GNU tar's ./configure test for getcwd()'s handling of excessively long path names. va_end() was found by manual inspection while hunting that down.
One of these (va_end) is cosmetic on my system but fixes API usage, while the other (findbranch) fixes an annoying crash on my machine.