summaryrefslogtreecommitdiff
path: root/doc/todo
diff options
context:
space:
mode:
Diffstat (limited to 'doc/todo')
-rw-r--r--doc/todo/Apt.trustsKey_should_not_invoke_apt-key.mdwn9
-rw-r--r--doc/todo/Apt.trustsKey_should_not_invoke_apt-key/comment_1_49003d4fdd0e75d477415cb0bb6bbd3c._comment8
-rw-r--r--doc/todo/apt_mark_support.mdwn27
-rw-r--r--doc/todo/spin_failure_HEAD.mdwn2
-rw-r--r--doc/todo/spin_failure_HEAD/comment_1_9c7d9ae7860d9cfc28e7d015b015dc2e._comment9
-rw-r--r--doc/todo/spin_failure_HEAD/comment_2_a9b7013305a7f8d58175510b57bbadd2._comment8
-rw-r--r--doc/todo/spin_failure_HEAD/comment_3_952939a1333d6fc24ed288a80b76f168._comment8
-rw-r--r--doc/todo/spin_failure_HEAD/comment_4_684adfe4d134b4e27ed00db62f8e3372._comment43
-rw-r--r--doc/todo/support_for_libvirt_KVM_VMs/comment_2_54538a03d7085513538baa2970983ae0._comment8
9 files changed, 121 insertions, 1 deletions
diff --git a/doc/todo/Apt.trustsKey_should_not_invoke_apt-key.mdwn b/doc/todo/Apt.trustsKey_should_not_invoke_apt-key.mdwn
new file mode 100644
index 00000000..d2f9e6db
--- /dev/null
+++ b/doc/todo/Apt.trustsKey_should_not_invoke_apt-key.mdwn
@@ -0,0 +1,9 @@
+[Over at the Debian BTS](https://bugs.debian.org/907290), David Bremner points out that the apt-key manpage, on Debian stable, says
+
+> Instead of using this [add] [sub]command a keyring should be placed
+> directly in the /etc/apt/trusted.gpg.d/ directory with a
+> descriptive name and either "gpg" or "asc" as file extension.
+
+So ISTM that `Apt.trustsKey` should be implemented simply with `File.hasContents`, i.e., the property should delete any old `.gpg` file and then create a `.asc` file with the text string content of the `AptKey`.
+
+--spwhitton
diff --git a/doc/todo/Apt.trustsKey_should_not_invoke_apt-key/comment_1_49003d4fdd0e75d477415cb0bb6bbd3c._comment b/doc/todo/Apt.trustsKey_should_not_invoke_apt-key/comment_1_49003d4fdd0e75d477415cb0bb6bbd3c._comment
new file mode 100644
index 00000000..67ad5260
--- /dev/null
+++ b/doc/todo/Apt.trustsKey_should_not_invoke_apt-key/comment_1_49003d4fdd0e75d477415cb0bb6bbd3c._comment
@@ -0,0 +1,8 @@
+[[!comment format=mdwn
+ username="joey"
+ subject="""comment 1"""
+ date="2018-09-17T18:16:56Z"
+ content="""
+Fine by me as long as it cleans up or overwrites the file that apt-key
+installed earlier.
+"""]]
diff --git a/doc/todo/apt_mark_support.mdwn b/doc/todo/apt_mark_support.mdwn
new file mode 100644
index 00000000..50591222
--- /dev/null
+++ b/doc/todo/apt_mark_support.mdwn
@@ -0,0 +1,27 @@
+I'd like a property that removes all packages that were not installed by
+the current set of propellor properties. For systems that are fully
+specified by propellor, this would keep the cruft from piling up.
+
+This could be done using apt-mark. Before propellor installs anything with
+apt, go through the apt-mark list and set all packages to auto. When apt is
+run to install a package, it will mark it as manually installed. Since
+Apt.installed skips running apt when packages are already installed, it
+would need to either be changed to run apt anyway, or to run apt-mark
+manual. And then after all other properties, run apt-get autoremove.
+
+Running the autoremove at the end is supported by the propellor monad,
+but there's currently no way to run something before all properties.
+The first Apt.install to run could handle the apt-mark-list-to-auto part,
+although there's also not currently any state for the property to keep
+track of if it's run before.
+
+It would also be possible to not do the apt-mark at the beginning. Instead,
+make the Propellor monad a Writer (polymorphized somehow perhaps like Info
+is) and have Apt.install track the packages that are installed. (Or it
+could be changed to a HasInfo property, and then the list of packages would
+accumulate in Info, but there are likely things that use Apt.installed
+inside ensureProperty which that would cause problems for.)
+
+Either way, an action run at the end can then update the apt-mark data to
+reflect the gathered list of packages, and run apt-get autoremove.
+--[[Joey]]
diff --git a/doc/todo/spin_failure_HEAD.mdwn b/doc/todo/spin_failure_HEAD.mdwn
index 1a591b35..e49df633 100644
--- a/doc/todo/spin_failure_HEAD.mdwn
+++ b/doc/todo/spin_failure_HEAD.mdwn
@@ -81,7 +81,7 @@ Sending privdata (73139 bytes) to kite.kitenet.net ... done
[2017-06-18 16:31:16 EDT] received marked GITPUSH
[2017-06-18 16:31:16 EDT] command line: GitPush 11 12
16:31:16.361717 pkt-line.c:80 packet: fetch< 17abde8439d17d49676f549f357f45eb2adce868 refs/remotes/db48x/master
-<pre>
+</pre>
> > So there's an actual protocol error here; the first 13 lines
> > of git protocol were not sent.
diff --git a/doc/todo/spin_failure_HEAD/comment_1_9c7d9ae7860d9cfc28e7d015b015dc2e._comment b/doc/todo/spin_failure_HEAD/comment_1_9c7d9ae7860d9cfc28e7d015b015dc2e._comment
new file mode 100644
index 00000000..8fb8a027
--- /dev/null
+++ b/doc/todo/spin_failure_HEAD/comment_1_9c7d9ae7860d9cfc28e7d015b015dc2e._comment
@@ -0,0 +1,9 @@
+[[!comment format=mdwn
+ username="david@1439a1cab13195a56248b6a8fd98a62028bcba8a"
+ nickname="david"
+ avatar="http://cdn.libravatar.org/avatar/22c2d800db6a7699139df604a67cb221"
+ subject="still in 5.3.6"
+ date="2018-08-24T02:12:44Z"
+ content="""
+I'm seeing this problem in 5.3.6, but only when the remote is Debian stable. Both ends are running 5.3.6 built from source.
+"""]]
diff --git a/doc/todo/spin_failure_HEAD/comment_2_a9b7013305a7f8d58175510b57bbadd2._comment b/doc/todo/spin_failure_HEAD/comment_2_a9b7013305a7f8d58175510b57bbadd2._comment
new file mode 100644
index 00000000..a8866294
--- /dev/null
+++ b/doc/todo/spin_failure_HEAD/comment_2_a9b7013305a7f8d58175510b57bbadd2._comment
@@ -0,0 +1,8 @@
+[[!comment format=mdwn
+ username="david"
+ avatar="http://cdn.libravatar.org/avatar/22c2d800db6a7699139df604a67cb221"
+ subject="still in 5.4.1, but only on one machine"
+ date="2018-08-24T10:11:16Z"
+ content="""
+I updated to 5.4.1, and I still consistenly see this trying to spin my office computer from home. Weirdly a VM running Debian stretch on the same network does not repropduce. I'll have to try from a different machine on the office network to see if that makes a difference.
+"""]]
diff --git a/doc/todo/spin_failure_HEAD/comment_3_952939a1333d6fc24ed288a80b76f168._comment b/doc/todo/spin_failure_HEAD/comment_3_952939a1333d6fc24ed288a80b76f168._comment
new file mode 100644
index 00000000..98d7f18b
--- /dev/null
+++ b/doc/todo/spin_failure_HEAD/comment_3_952939a1333d6fc24ed288a80b76f168._comment
@@ -0,0 +1,8 @@
+[[!comment format=mdwn
+ username="david"
+ avatar="http://cdn.libravatar.org/avatar/22c2d800db6a7699139df604a67cb221"
+ subject="definitely network related"
+ date="2018-08-24T13:58:49Z"
+ content="""
+I can spin the same host from a different host on the office LAN (in fact they are connected to a cheapo hub, so that might not be much of a test), and from itself. So I guess it definitely has to do with networking. Does propellor need anything other than port 22 open?
+"""]]
diff --git a/doc/todo/spin_failure_HEAD/comment_4_684adfe4d134b4e27ed00db62f8e3372._comment b/doc/todo/spin_failure_HEAD/comment_4_684adfe4d134b4e27ed00db62f8e3372._comment
new file mode 100644
index 00000000..3f6aebcb
--- /dev/null
+++ b/doc/todo/spin_failure_HEAD/comment_4_684adfe4d134b4e27ed00db62f8e3372._comment
@@ -0,0 +1,43 @@
+[[!comment format=mdwn
+ username="joey"
+ subject="""comment 4"""
+ date="2018-08-26T14:50:22Z"
+ content="""
+As far as I know, this was fixed in a series of commits,
+[[!commit 01fc1375cece096ab2dec480b843ecdbc4f0d94e]]
+[[!commit 1555c6f88a0446d3e29149eff8315817696731e1]]
+[[!commit 53fe5ffaac4a243bb9fd3cf0e757128150a6a199]]
+
+The problem was intermittent for me, I think based on network timing and
+different buffering behavior with different timings,
+which made it hard to debug, but I've not seen it since and I was seeing it
+frequently enough to be fairly sure I fixed it.
+
+So I wonder if you might have some sort of version skew issue on the host
+being spun (eg, it could have an old version of propellor installed and be
+failing before spin can update it to the fixed version). The easy way to
+verify you have the fixed version is to run `git config propellor.debug 1`
+in /usr/local/propellor/ on the host being spun, and look for
+"--upload-pack ./propellor --gitpush" in a debug message.
+
+If you're confident the remote propellor has the above commits in it,
+you're going to need to do some debugging.
+Setting `GIT_TRACE_PACKET=1` on the remote system was very helpful in
+understanding the problem, and should probably be your first step.
+Setting that environment inside Propellor.Spin.gitPullFromUpdateServer
+should work, of course you'll have to get the remote host to build
+propellor with that change somehow despite --spin to it not working.
+
+ diff --git a/src/Propellor/Spin.hs b/src/Propellor/Spin.hs
+ index 4a945e82..aa73e3b7 100644
+ --- a/src/Propellor/Spin.hs
+ +++ b/src/Propellor/Spin.hs
+ @@ -359,6 +359,7 @@ spinCommitMessage = "propellor spin"
+ -- to receive the data.
+ gitPullFromUpdateServer :: IO ()
+ gitPullFromUpdateServer = req NeedGitPush gitPushMarker $ \_ -> do
+ + setEnv "GIT_TRACE_PACKET" "1" True
+ -- IO involving stdin can cause data to be buffered in the Handle
+ -- (even when it's set NoBuffering), but we need to pass a FD to
+ -- git fetch containing all of stdin after the gitPushMarker,
+"""]]
diff --git a/doc/todo/support_for_libvirt_KVM_VMs/comment_2_54538a03d7085513538baa2970983ae0._comment b/doc/todo/support_for_libvirt_KVM_VMs/comment_2_54538a03d7085513538baa2970983ae0._comment
new file mode 100644
index 00000000..497e364a
--- /dev/null
+++ b/doc/todo/support_for_libvirt_KVM_VMs/comment_2_54538a03d7085513538baa2970983ae0._comment
@@ -0,0 +1,8 @@
+[[!comment format=mdwn
+ username="joey"
+ subject="""comment 2"""
+ date="2018-10-14T00:36:55Z"
+ content="""
+@david, but you'd not then want to change the backing raw image, I assume,
+or does qcow somehow deal with that?
+"""]]