workingctx: correctly compute the flag for noexec filesystems+merge
This bug happens if the filesystem doesn't support exec-bit, during merges,
for example in 24ed7a541f23 on the hg repo.
If f is not in p1, but is in p2 and has the x-bit in p2, since the dirstate is
based on p1, and the FS doesn't support the exec-bit, the dirstate can't
"guess" the right bit.
We instead fix it in workingcontext.flags()/manifest.
#!/bin/sh
# test that we don't interrupt the merge session if
# a file-level merge failed
hg init repo
cd repo
echo foo > foo
echo a > bar
hg ci -Am 'add foo'
hg mv foo baz
echo b >> bar
echo quux > quux1
hg ci -Am 'mv foo baz'
hg up -qC 0
echo >> foo
echo c >> bar
echo quux > quux2
hg ci -Am 'change foo'
# test with the rename on the remote side
HGMERGE=false hg merge
hg resolve -l
# test with the rename on the local side
hg up -C 1
HGMERGE=false hg merge
echo % show unresolved
hg resolve -l
echo % unmark baz
hg resolve -u baz
echo % show
hg resolve -l
hg st
echo % re-resolve baz
hg resolve baz
echo % after
hg resolve -l
echo % resolve all warning
hg resolve
echo % resolve all
hg resolve -a
echo % after
hg resolve -l
true