[Yum-devel] [PATCH] use reget=check_timestamp when downloading repomd.xml

[James Antill]

On Fri, 2012-06-22 at 13:40 +0200, Zdeněk Pavlas wrote:
> This is not significantly faster than always downloading the
> whole file (repomd.xml is typically about 3kB), we just save
> the bandwidth.
> 
> With reget on, urlgrabber does not touch the local file
> on (most of) failures hence no reverting is needed.
> 
> Won't work with older urlgrabber which raises NotImplemented.
> Needs to bump Requires: in .spec

 You _really_ want to find a way to test this at runtime or something.

> ---
>  yum/yumRepo.py |   20 ++------------------
>  1 files changed, 2 insertions(+), 18 deletions(-)
> 
> diff --git a/yum/yumRepo.py b/yum/yumRepo.py
> index 14f6d5f..2d8a1ba 100644
> --- a/yum/yumRepo.py
> +++ b/yum/yumRepo.py
> @@ -1090,41 +1090,25 @@ Insufficient space in download directory %s
>          checkfunc = (self._checkRepoXML, (), {})
>          if grab_can_fail is None:
>              grab_can_fail = 'old_repo_XML' in self._oldRepoMDData
> -        tfname = ''
>          try:
> -            # This is named so that "yum clean metadata" picks it up
> -            tfname = tempfile.mktemp(prefix='repomd', suffix="tmp.xml",
> -                                     dir=os.path.dirname(local))

 You can't just delete this ... not using temporary files breaks all the
revert logic.