summaryrefslogtreecommitdiffstats
path: root/TODO
diff options
context:
space:
mode:
authorDaniel Baumann <daniel.baumann@progress-linux.org>2024-04-20 16:34:05 +0000
committerDaniel Baumann <daniel.baumann@progress-linux.org>2024-04-20 16:34:05 +0000
commit17910b92b762000663c665fedbab30f7b1d41bdf (patch)
tree2e8c4fbab094dac1e376dc9395e3cf1cc5e40639 /TODO
parentInitial commit. (diff)
downloadusrmerge-upstream/39.tar.xz
usrmerge-upstream/39.zip
Adding upstream version 39.upstream/39upstream
Signed-off-by: Daniel Baumann <daniel.baumann@progress-linux.org>
Diffstat (limited to '')
-rw-r--r--TODO27
1 files changed, 27 insertions, 0 deletions
diff --git a/TODO b/TODO
new file mode 100644
index 0000000..5e76bea
--- /dev/null
+++ b/TODO
@@ -0,0 +1,27 @@
+How to implement race-free replacement of a directory by a symlink:
+
+mount -o bind / /tmp/root
+
+mount -o bind /usr/bin /bin
+mv /tmp/root/bin /tmp/root/bin.old
+ln -s usr/bin /tmp/root/bin
+umount /bin
+rm -rf /tmp/root/bin.old
+
+umount /tmp/root
+
+Is this complexity justified just for convert_directory()?
+
+For some operations there are two possible implementations:
+- cp/rename/symlink: slower but race-free
+- rename/symlink: faster (only metadata operations, as long as / and /usr
+ are on the same file system) but racy
+
+Is it useful to keep the first implementation if we do not also fix
+the directory-to-symlink races?
+
+How to handle the initramfs check in preinst?
+Is asking a debconf question justified if we suspect that there is no
+initramfs?
+Or should we always ask for confirmation if /usr is a standalone filesystem?
+