Libguestfs - Mar 2019 - Re: [supermin PATCH] rebuild the output it when SUPERMIN_KERNEL or SUPERMIN_MODULES are defined

SUPERMIN_KERNEL and SUPERMIN_MODULES don't work to guestfish.

Since guestfish sets --if-newer parameter to supermin, so the environment
variables are not used under the following conditions.
  - the output directory exists and,
  - the dates of both input files and package database are
    older than the output

To solve that, rebuild the output it when SUPERMIN_KERNEL or
SUPERMIN_MODULES are defined even if --if-newer is set.

Signed-off-by: Masayoshi Mizuma <msys.mizuma at gmail.com>
---
 src/supermin.ml | 18 +++++++++++-------
 1 file changed, 11 insertions(+), 7 deletions(-)

diff --git a/src/supermin.ml b/src/supermin.ml
index 7c7135b3..b997643 100644
--- a/src/supermin.ml
+++ b/src/supermin.ml
@@ -236,13 +236,17 @@ appliance automatically.
    *)
   if if_newer then (
     try
-      let odate = (lstat outputdir).st_mtime in
-      let idates = List.map (fun d -> (lstat d).st_mtime) inputs in
-      let pdate = (get_package_handler ()).ph_get_package_database_mtime () in
-      if List.for_all (fun idate -> idate < odate) (pdate :: idates) then
(
-        if debug >= 1 then
-          printf "supermin: if-newer: output does not need
rebuilding\n%!";
-        exit 0
+      let kernel = try getenv "SUPERMIN_KERNEL" with Not_found ->
"" in
+      let modules = try getenv "SUPERMIN_MODULES" with Not_found
-> "" in
+      if kernel = "" && modules = "" then (
+        let odate = (lstat outputdir).st_mtime in
+        let idates = List.map (fun d -> (lstat d).st_mtime) inputs in
+        let pdate = (get_package_handler ()).ph_get_package_database_mtime ()
in
+        if List.for_all (fun idate -> idate < odate) (pdate :: idates)
then (
+          if debug >= 1 then
+            printf "supermin: if-newer: output does not need
rebuilding\n%!";
+          exit 0
+        )
       )
     with
       Unix_error _ -> () (* just continue *)
-- 
2.20.1

On Monday, 4 March 2019 12:18:58 CET Masayoshi Mizuma
wrote:
> SUPERMIN_KERNEL and SUPERMIN_MODULES don't work to guestfish.
> 
> Since guestfish sets --if-newer parameter to supermin, so the environment
> variables are not used under the following conditions.
>   - the output directory exists and,
>   - the dates of both input files and package database are
>     older than the output
> 
> To solve that, rebuild the output it when SUPERMIN_KERNEL or
> SUPERMIN_MODULES are defined even if --if-newer is set.
> 
> Signed-off-by: Masayoshi Mizuma <msys.mizuma at gmail.com>
> ---
This approach switches from one side of the situation
(= SUPERMIN_KERNEL and SUPERMIN_MODULES are not taken into account by
--if-newer) to the very opposite side (= setting them always rebuild
the appliance).  I do not think this is a good idea, since the current
situation is easy to workaround (`rm -rf $(guestfish get-cachedir)`),
while using a different kernel will always rebuild the appliance after
this change (and thus slow every run down).

What is the goal here? Make sure that --if-newer actually rebuilds an
appliance when changing the values of SUPERMIN_KERNEL and
SUPERMIN_MODULES (even setting/unsetting them)?

I agree that supermin ought to do better in --if-newer checks: for
example, removing any of the files of a ext2 appliance (e.g. "root",
or "kernel") will not make --if-newer rebuild the appliance.
Maybe a better idea could be to record files/timestamps for appliances,
so supermin can easily check what's missing/older.

-- 
Pino Toscano
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 833 bytes
Desc: This is a digitally signed message part.
URL:
<http://listman.redhat.com/archives/libguestfs/attachments/20190304/9890e380/attachment.sig>

On Mon, Mar 04, 2019 at 01:17:26PM +0100, Pino Toscano
wrote:
> On Monday, 4 March 2019 12:18:58 CET Masayoshi Mizuma wrote:
> > SUPERMIN_KERNEL and SUPERMIN_MODULES don't work to guestfish.
> > 
> > Since guestfish sets --if-newer parameter to supermin, so the
environment
> > variables are not used under the following conditions.
> >   - the output directory exists and,
> >   - the dates of both input files and package database are
> >     older than the output
> > 
> > To solve that, rebuild the output it when SUPERMIN_KERNEL or
> > SUPERMIN_MODULES are defined even if --if-newer is set.
> > 
> > Signed-off-by: Masayoshi Mizuma <msys.mizuma@gmail.com>
> > ---
> 
> This approach switches from one side of the situation
> (= SUPERMIN_KERNEL and SUPERMIN_MODULES are not taken into account by
> --if-newer) to the very opposite side (= setting them always rebuild
> the appliance).  I do not think this is a good idea, since the current
> situation is easy to workaround (`rm -rf $(guestfish get-cachedir)`),
> while using a different kernel will always rebuild the appliance after
> this change (and thus slow every run down).
Thanks. I didn't know the workaround.
> 
> What is the goal here? Make sure that --if-newer actually rebuilds an
> appliance when changing the values of SUPERMIN_KERNEL and
> SUPERMIN_MODULES (even setting/unsetting them)?
Yes, that is my goal.
> I agree that supermin ought to do better in --if-newer checks: for
> example, removing any of the files of a ext2 appliance (e.g.
"root",
> or "kernel") will not make --if-newer rebuild the appliance.
> Maybe a better idea could be to record files/timestamps for appliances,
> so supermin can easily check what's missing/older.
It's great idea, thanks!
I'll try to implement that.

- Masa