On Jun 4, 2010, at 5:25 AM, <raino.lintula...@elektrobit.com> wrote:
> Hi Bob, > here are few issues which I've noticed while trying buckets feature. > I'm not sure whether all of them are bugs or have I done something > wrong with bucket pool and bucket definitions. I may have also > interpreted wrongly how buckets should be used. > > br Raino > ------------------------------------------------------ > FOSSology svn tar ball downloaded June 3rd. > > Issue#1. Bucket browser does not show Count correctly > Procedure: > a. run bucket analysis for package > b. Browse --> Bucket browser for the package > c. Count = 1 > Expected result: > Count is the number of files detected by the defined bucket Yes, this is confusing about buckets. Just so others understand what you are talking about, here is a bucket report: http://fossology.org/~bobg/buckets_6.4.10/bucket1.png and if you click "Show" by "No licenses found", you might expect to see 4 files in this bucket. Instead you see 7: http://fossology.org/~bobg/buckets_6.4.10/bucketlist.png What is happening is that the bucket browser (bucket1.png) is showing you which buckets the files listed in the file list are in. If you click on "No licenses found", the bucket browser will highlight the files on the right that are in this bucket. http://fossology.org/~bobg/buckets_6.4.10/bucket2.png So you can see the four files on the right that are in this bucket. And if you add the counts you get 7 which equals the number of files on the right. The confusing thing is that the bucketlist (bucketlist.png) shows you ALL the files that are in the given bucket, not just which files in the bucket browser file list are in this bucket. So if, in the right hand file list, you have a container (directory, rpm, tar, etc), the file list (bucketlist.png) will also show all those individual files inside that container. The reason I don't make the Count in the browser equal to the number of individual files in that bucket is that, unlike licenses, containers can be in a bucket AND buckets are hierarchical. For example, say you have the following: 1) a tar file, mytar.tar that contains two files, a.c and b.c. 2) you have defined two buckets "GPL v3" and "All BSD variations" 3) a.c is in the gpl bucket, b.c is in the bsd bucket 3) you have defined the gpl bucket to be a higher priority (bucket_evalorder) than the other. If you point the bucket browser at mytar.tar (showing only mytar.tar in the right hand file list), then it will show Count=1 for bucket GPL v3. In other words the container mytar.tar is in the gpl bucket. But if you drill down into mytar.tar, you will see a.c and b.c in the file list and you will see Count=1 for GPL bucket, and Count=1 for the BSD bucket. We do hierarchical buckets to try and streamline the workflow for an IP attorney looking at the results, so they see the most important bucket for the given package/container. I'm always open to ideas for how to make this simpler. > Issue#2. Bucket browser does not show paged file list of bucket results > correctly > Procedure: > a. same first two steps as in previous case > b. select Show in files column > c. The list of files in bucket is shown. > d. Select Next (or any page number) to show next files in bucket > e. File list is not changing (page 1 files are constantly shown) > Expected result: > Next page of file list is shown. > However, if 'Show All' is selected then files are shown correctly. Thanks. I'll put fixing this on the list. > Issue#3.Bucket analysis with new Bucket pool version cannot be done for > the package > for which older version bucket analysis was done > Procedure: > a. Change version in bucketpool > b. select Jobs-->Agents > c. select upload for which bucket analysis was done > d. Bucket analysis is not show on 'Select additional analysis' list > Expected result: > Because bucketpool version was changed then bucket analysis > should be able to scheduled again for the uploaded package The Jobs->Agents UI needs to be updated. However, you don't want to change the version in a bucketpool record (I assume you also changed bucket definitions - bucketdef records). What you will need to do is to define a new set (a new bucketpool record with the new version, and new bucketdef records that refer to the new bucket pool). This way when you requeue the job (once the UI is updated) you will keep your old bucket results and create new results for the new bucket definitions. This allows you to see what effect your change had and it allows old url's, which refer to the old bucketpool, to still show the same results as before. We are trying to keep url's persistent. I should probably write a tool to clone bucketpools to make them easier to modify. If you want a work around so you can requeue buckets, and you don't care about persistent url's, do this: 1) Jobs > Queue > by Upload Click on "Delete" for the bucket job. 2) delete the record for this upload in table bucket_ars. 3) change the bucketpool_pk and all the bucketdef records that point to it. Or clone the bucketpool and bucketdef records as mentioned above. Then you can go into Jobs>Agents and requeue. > Issue#4.Bucket analysis with new Bucket pool cannot be done for the > package > for which bucket analysis based on different bucket pool was done > earlier > Procedure: > a. Define and add a new bucketpool > b. Change the default bucket pool in user account to a new bucket pool > b. select Jobs-->Agents > c. select upload for which bucket analysis was done with the old bucket > pool > d. Bucket analysis is not show on 'Select additional analysis' list > Expected result: > Because a different bucketpool was selected to be used then bucket > analysis > should be able to scheduled for the uploaded package This is the same problem as #3 where the UI see's that buckets were already run and won't let you reschedule. The work around in this case is to 1) Jobs > Queue > by Upload Click on "Delete" for the bucket job. Thanks for your comments. Suggestions are always welcome on how this should work and what tools you need. Bob Gobeille b...@fossology.org _______________________________________________ fossology mailing list fossology@fossology.org http://fossology.org/mailman/listinfo/fossology
