setup: run require_auto_deps() before attempting to import any deps in __init__.py
For one thing, this makes missing-dependency failures into DistributionNotFound errors instead of ImportErrors, which might be more useful to the user. For another thing, if someone is using distributions that were installed with --multi-version, then they might be not importable until after require_auto_deps() has been run. (The docs claim that this would be the case, but we don't have an example of this happening at this time.)
setup: show-tool-versions: emit module and __version__ information even when module name != distribution (package) name, and add TwistedCore, TwistedWeb, and TwistedConch
david-sarah [Fri, 29 Oct 2010 23:53:28 +0000 (16:53 -0700)]
scripts/runner.py: remove pkg_resources.require() calls. These are at best redundant because we have already called _auto_deps.require_auto_deps() (from allmydata.__init__) at that point, and they are causing failure of the test-from-prefixdir step on some buildslaves. refs #1190
david-sarah [Fri, 29 Oct 2010 22:31:11 +0000 (15:31 -0700)]
bundled zetuptoolz: if __main__.__requires__ exists then do not add packages to the working set if they provide an incompatible version of a package. Also put a complete __requires__ listing the transitive closure of dependencies at the beginning of generated scripts, rather than a shallow __requires__ specifying only the application version. refs #1190
david-sarah [Fri, 29 Oct 2010 22:28:25 +0000 (15:28 -0700)]
setup.py, misc/build_helpers/run_trial.py: use undocumented __requires__ variable to cause setuptools/zetuptoolz to put the correct versions of dependencies on sys.path. Also ensure that run_trial adds the bundled zetuptoolz egg at the start of sys.path if present. Make the source directory comparison work correctly for the test-with-fake-pkg build step. refs #1190
david-sarah [Fri, 29 Oct 2010 19:18:04 +0000 (12:18 -0700)]
test_runner.py: if the only noise is 'UserWarning: Unbuilt egg for setuptools', skip instead of failing the no_noise tests. This version avoids 'any' to be compatible with Python < 2.5. refs #1235.
david-sarah [Thu, 28 Oct 2010 18:00:46 +0000 (11:00 -0700)]
NEWS: clarify (strengthen) description of what backdoors.rst declares, and add bugfix entries for 'tahoe cp' and Windows console bugs. refs #1216, #1224, #1232
david-sarah [Wed, 27 Oct 2010 04:33:02 +0000 (21:33 -0700)]
make ResponseCache smarter to avoid memory leaks: don't record timestamps, use DataSpans to merge entries, and clear the cache when we see a new seqnum. refs #1045, #1229
tahoe_cp.py: Don't call urllib.quote with an Unicode argument, fix #1224
tahoe_backup.py: Fix another (potential) occurrence of calling urllib.quote()
with an Unicode parameter
Brian Warner [Fri, 29 Oct 2010 08:20:36 +0000 (01:20 -0700)]
fix #1223, crash+inefficiency during repair due to read overrun
* repairer (really the uploader) reads beyond end of input file (Uploadable)
* new-downloader does not tolerate overreads
* uploader does lots of tiny reads (inefficient)
This fixes the last two. The uploader still does a single overread at the end
of the input file, but now that's ok so we can leave it in place. The
uploader now expects the Uploadable to behave like a normal disk
file (reading beyond EOF will return less data than was asked for), and now
the new-downloadable behaves that way.
Brian Warner [Wed, 27 Oct 2010 06:13:42 +0000 (23:13 -0700)]
'tahoe start': use import+call rather than find+spawn
This removes the need to use a locally-built (dependency) bin/twistd, and
removes a big chunk of behavior differences between unix and windows. It
also happens to resolve the "client node probably started" uncertainty.
Might help with #1190, #602, and #71.
setup: catch and log ValueError from locale.getdefaultlocale() in show-tool-versions.py
I got a bug report from Mathias Baert showing that locale.getdefaultlocale() raises an exception on his Mac OS X system. Heh.
setup: add --multi-version to the "setup.py develop" command-line
fixes #530. I earlier tried this twice (see #530 for history) and then twice rolled it back due to some problems that arose. However, I didn't write down what the problems were in enough detail on the ticket that I can tell today whether those problems are still issues, so here goes the third attempt. (I did write down on the ticket that it would not create site.py or .pth files in the target directory with --multi-version mode, but I didn't explain why *that* was a problem.)
setup: allow an older version of pycryptopp if we are not on x86 or amd64 architecture (which are the architectures on which older versions of Crypto++ had bugs in asm implementations of AES and SHA-256)
test: make tests stop relying on pyutil version class accepting the string 'unknown' for its version, and make them forward-compatible with the future Python Rational Version Numbering standard
Brian Warner [Fri, 10 Sep 2010 04:30:38 +0000 (21:30 -0700)]
immutable download: have the finder inform its share consumer "no more shares" in a subsequent tick, thus avoiding accidentally telling it "no more shares" now and then telling it "here's another share" in a subsequent tick
fixes #1191
Patch by Brian. This patch description was actually written by Zooko, but I forged Brian's name on the "author" field so that he would get credit for this patch in revision control history.
immutable downloader: add a test specifically of whether the finder sometimes announces "no more shares ever" and then announces a newly found share
(The current code fails this test, ref #1191.)
tests: assign the storage servers to a fixed order which triggers a bug in new downloader every time this test is run (formerly this test would detect the bug in new-downloader only sporadically)
If you are investigating the bug in new-downloader, one way to investigate might be to change this ordering to a different fixed order (e.g. rotate by 4 instead of rotate by 5) and observe how the behavior of new-downloader differs in that case.
tests: bump up the allowed number of reads
Kyle's OpenBSD buildslave used 41 reads when doing this test. The fact that I'm blindly bumping this number up to match the observed behavior probably means this isn't a good criterion to be testing for anyway. But perhaps someone else (Brian) could investigate why that run on Kyle's OpenBSD box took four more reads than we expected, and whether the fact that it took 41 reads to do this operation is indicative of an actual problem.
Brian Warner [Wed, 1 Sep 2010 01:37:02 +0000 (18:37 -0700)]
SegmentFetcher: use new diversity-seeking share-selection algorithm, and
deliver all shares at once instead of feeding them out one-at-a-time.
Also fix distribution of real-number-of-segments information: now all
CommonShares (not just the ones used for the first segment) get a
correctly-sized hashtree. Previously, the late ones might not, which would
make them crash and get dropped (causing the download to fail if the initial
set were insufficient, perhaps because one of their servers went away).
Update tests, add some TODO notes, improve variable names and comments.
Improve logging: add logparents, set more appropriate levels.
Brian Warner [Mon, 9 Aug 2010 22:49:09 +0000 (15:49 -0700)]
immutable/filenode.py: put off DownloadStatus creation until first read() call
This avoids spamming the "recent uploads and downloads" /status page from
FileNode instances that were created for a directory read but which nobody is
ever going to read from. I also cleaned up the way DownloadStatus instances
are made to only ever do it in the CiphertextFileNode, not in the
higher-level plaintext FileNode. Also fixed DownloadStatus handling of read
size, thanks to David-Sarah for the catch.
projects / tahoe-lafs / tahoe-lafs.git / log