On Fri, 4 Apr 2008, Peter Stuge wrote:
On Fri, Apr 04, 2008 at 02:25:31PM +0200, Uwe Hermann wrote:
On Fri, Apr 04, 2008 at 10:56:37AM +0200, Peter Stuge wrote:
I really like this idea but the patch depends on svn being installed to build properly.
Yes, that was (partially) discussed [1] in the thread leading up to the original implementation for superiotool.
Yes, but I think it would be ok if we use "unknown" or "0" or something like that if svn is not available
For tarballs that we make available I don't like that so much. It is very useful information to have in debug logs.
True, that would be nice. However it is non-trivial to automatically get a more *accurate* (fallback) revision number in a non-subversion environment.
The needed information is lost when the tarball is generated, e.g. by viewvc. Extracting and preserving that information on the fly looks cumbersome (would neccessitate changes to viewvc and whatever generates the snapshots at qa.coreboot.org)
OTOH, using a not-so-accurate fallback like "unknown" or "0" (possibly even the revision number of the Makefile) looks perfectly tractable: substitute the fallback when the output from the svnversion pipeline is empty.
(haven't tried what happens currently).
I expect it to just be blank.
Yes (I have seen some superiotool dumps like that on the list).
Also note that the patch at the start of this thread is based on the superiotool/flashrom code, which does not recurse into subdirectories [2]. A subdirectory aware pipeline is available in [3].
[1] http://www.coreboot.org/pipermail/coreboot/2007-October/025606.html [2] http://www.coreboot.org/pipermail/coreboot/2007-October/025630.html [3] http://www.coreboot.org/pipermail/coreboot/2007-October/026139.html
/ulf