[PATCH] auto-update-debuginfo: don't update cache when -C. BZ 1002449

[James Antill]

 Sorry, I did look at this and meant to reply but must have not marked
it as unread again ... anyway.

On Thu, 2013-08-29 at 12:06 +0200, Zdenek Pavlas wrote:
> - _write_cached() should be probably avoided when cacheonly.

 Maybe, I thought we were changing so that cacheonly just meant "don't
talk over the network".

> - when we can't cache the result, we should not run the slow
>   len(returnPackages) path at all.

 Why? That's the only way to get the correct result, if we don't have it
cached.

> - the cached num, even if stale, is better than assuming num==0.

 Sure, but we can just work it out and get the correct result.

> ---
>  plugins/auto-update-debuginfo/auto-update-debuginfo.py | 4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/plugins/auto-update-debuginfo/auto-update-debuginfo.py b/plugins/auto-update-debuginfo/auto-update-debuginfo.py
> index 539dc91..c0fa85a 100644
> --- a/plugins/auto-update-debuginfo/auto-update-debuginfo.py
> +++ b/plugins/auto-update-debuginfo/auto-update-debuginfo.py
> @@ -103,12 +103,12 @@ def prereposetup_hook(conduit):
>      if caching:
>          cfname = yb.conf.persistdir + '/plugins/auto-update-debuginfo/num'
>          crpmdbv, num = _read_cached(cfname)
> -        if num is not None:
> +        if num is not None and not yb.conf.cache:
>              rpmdbv = yb.rpmdb.simpleVersion(main_only=True)[0]
>              if rpmdbv != crpmdbv:
>                  num = None

 What is the logic here? If we have the correctly cached result why does
it matter if we are in cacheonly mode or not?


> -    if num is None:
> +    if num is None and not yb.conf.cache:
>          num = len(yb.rpmdb.returnPackages(patterns=['*-debuginfo']))
>          if caching:
>              rpmdbv = yb.rpmdb.simpleVersion(main_only=True)[0]

 This does make things faster ... but will give the wrong result. If we
really need to make this part faster we could maybe change the caching
to cache the *-debuginfo package names that we've seen before ... then
we don't really need much of a caching validation mechanism for the
exists case as we can just lookup the packages by name (almost instant
op.) and as soon as we find one we are good.