~racb/simplestreams/optimise-read

~racb/simplestreams/optimise-read_signed » Changes from revision 430

From Revision 430 to 411

Rev	Summary	Authors	Date
430	Optimise read_signed Optimise read_signed Use join instead of iteratively "appending" to a string. In my use case, this reduces signature check time from over 400 seconds to almost a tenth of a second.	Robie Basak	7 years ago
429	set a default user-agent set a default user-agent Follow on to the custom user agent merge by adding a default user agent.	Scott Moser	7 years ago
428	glance mirror: do not require that hypervisor_config be pres... glance mirror: do not require that hypervisor_config be present This just allows for hypervisor_config to not be present.	Scott Moser	7 years ago
427	Add user-agent support Add user-agent support This adds user-agent support. The library user may want to provide a specific user agent when querying a remote mirror. Because of the code structure, there's some not-very-nice code in here, mainly the use of url_reader_factory function, but without knowing what parts of simplestreams API we want to keep backward compatible, I refrained from changing anything of the existing code too much. There are tests for all the new code, and for some of the old code that was untested.	Scott Moser	7 years ago
426	Switch to using virt key for hypervisor type in stream data. Switch to using virt key for hypervisor type in stream data. In order to be consistent with other stream data that uses the 'virt' tag name (such as aws) the glance mirror should translate the underlying hypervisor_type for the glance image into an appropriate 'virt' value for the cloud. Also, change the following virt types lxc -> lxd qemu -> kvm	Scott Moser	8 years ago
425	Set product aliases without modifying release data dictionar... Set product aliases without modifying release data dictionary In r424 we added support for aliases. With that code it is possible that the daily devel products will lack the intended 'devel' alias. Aliases for a release/arch are set from toolutil.REL2VER[release] and if the stream is not daily the 'devel' item in the list of aliases is removed, however the code from that commit accidentally removed 'devel' from the source dictionary rather than creating a local copy first. If the devel release is evaluated before the devel daily then the 'devel' alias will not be present in the dictionary. This patch will copy the aliases from the release dictionary and modify the copy for the product being indexed.	Ben Howard	8 years ago
424	[MERGE] Added alias' to support LXD [MERGE] Added alias' to support LXD	Ben Howard	8 years ago
423	Enable better support for multi-hypervisor regions in OpenSt... Enable better support for multi-hypervisor regions in OpenStack 1) Determine correct disk format. 2) Set hypervisor_type on images based on ftype field.	Scott Moser	8 years ago
422	dump_data: dump json data with sorted keys dump_data: dump json data with sorted keys make dump_data use sorted keys for dictionaries. The desire is to reduce differences in output that would otherwise be the same.	Scott Moser	8 years ago
421	tools/make-test-data: Fix missing lxc/lxd files for ppc64el ... tools/make-test-data: Fix missing lxc/lxd files for ppc64el builds. We had a whitelist on ppc64el ever since the first ppc64el builds. Removing that whitelist means we will now get correct file information for all ppc64el builds.	Scott Moser	8 years ago
420	Fix TestUrlContentSourcea to test with requests, add build-d... Fix TestUrlContentSourcea to test with requests, add build-depends Fix the test case to use Requests, then fix the build failure that actually testing it caused. Then, support skipping the requests tests if it is not present.	Scott Moser	8 years ago
419	sanitize size parameter to read on Readers sanitize size parameter to read on Readers This fixes the read() calls in Urllib2UrlReader, RequestsUrlReader, and the new FileReader so that they all support: read(-1) read(None) read() and pass the correct value down to the python below it. Essentially, python3 prefers read(size=None) while python2 prefers read(size-1) if the size is passed explicitly rather than using read().	Scott Moser	8 years ago
418	fix lintian warnings, and apply pull back distro package bui... fix lintian warnings, and apply pull back distro package build changes debian/control: - bump standards version - add build-depends on python-mock, python-yamnl, python3-mock - add build-depends on dh-python per sbuild output recommendation debian/copyright: fix license string agplv3 to AGPL-3 debian/rules: set SS_REQUIRE_DISTRO_INFO = 0 during build	Scott Moser	8 years ago
417	fix a bug where --no-verify doesn't work with signed streams... fix a bug where --no-verify doesn't work with signed streams. Although the assignment of cmd was done only if checked was True, cmd was used when checked was not True. This produced a stack trace with UnboundLocalError. This branch fixes it and introduces tests based on test_signed_data.	Scott Moser	8 years ago
416	tools: xenial has .tar.gz files for the moment. tools: xenial has .tar.gz files for the moment. .tar.gz are expected to go away in xenial, but are present for the moment. As other things (maas image build process) needs them for now, we should index them.	Scott Moser	8 years ago
415	Only condense tags that are present in all entries. Only condense tags that are present in all entries. There is no way to indicate, at an item level, that a key is not present. This means that in order to be a candidate for compaction, the key must be present in all items. So using set.update() is not correct. We don't want the union of all keys in all items, we want the intersection of those keys.	Scott Moser	8 years ago
414	add json2streams command add json2streams command json2streams accepts a list of JSON files as input, each of those files containing a JSON list of simplestreams "items", as emitted by sstream-query --json. With --juju-format specified, it uses the Juju conventions for writing files (no colons, complete index in index2.json). Otherwise, it uses the standard simplestreams format. As a driveby: * write_streams to not modify its input. * write_streams to accept condense=True argument indicating if any condensing should be done in the data written.	Scott Moser	8 years ago
413	tools/ : xenial will not ship .tar.gz files tools/ : xenial will not ship .tar.gz files The .tar.gz file (compressed sparse tarball of unpartitioned disk image) will not ship in xenial. Modify the is_expected to address that.	Scott Moser	8 years ago
412	tools/toolutil.py: document is_expected much better tools/toolutil.py: document is_expected much better When working on is_expected, I was confused. This change is only: a.) a name change of 'repl' to 'suffix' b.) add description of the function	Scott Moser	8 years ago
411	tools/ : adjust to handle changes in /query data in xenial tools/ : adjust to handle changes in /query data in xenial xenial will not have a -root.tar.gz file, so it wont advertise that file in /query. Adjust the loading of download data to find the right one.	Scott Moser	8 years ago

Older »