summaryrefslogtreecommitdiff
Commit message (Collapse)AuthorAge
...
* Test suite: t-expect-push-fail checks HEAD object did not arriveIan Jackson2015-06-10
|
* Test suite: test forbidden objects not presentIan Jackson2015-06-10
|
* Test suite: Use t-expect-push-fail everywhereIan Jackson2015-06-10
|
* Test suite: Make t-expect-push-fail out of the remains of mustfailIan Jackson2015-06-10
|
* Test suite: Move most of mustfail into t-expect-failIan Jackson2015-06-10
|
* Test suite: Make t-expect-fail flexible wrt which grep to runIan Jackson2015-06-10
| | | | | | | | Change the default to fgrep, and override it in the places where this is needed. Also use shell substitution to quote .s in tag name in debpolicy-newreject.
* Test suite: Make t-reporefs stash the outputIan Jackson2015-06-10
|
* Test suite: drs-push-rejects: use set +o pipefail, not set +eIan Jackson2015-06-07
|
* Test suite: Move t-reporefs into libIan Jackson2015-06-07
|
* Test suite: set -o pipefail, in libIan Jackson2015-06-07
|
* Test suite: Prepare for pipefail in t-git-get-refIan Jackson2015-06-07
|
* Test suite: Prepare for pipefail in t-apply-diffIan Jackson2015-06-07
|
* Test suite: t-git-get-ref: check that argument starts with ref/Ian Jackson2015-06-07
|
* Test suite: Fix call to t-refs-notexist to have refs/Ian Jackson2015-06-07
|
* Test suite: Fully test replay attack preventionIan Jackson2015-05-31
|
* Test suite: Make it possible to call t-policy more than once (moving t-dsd ↵Ian Jackson2015-05-31
| | | | out of it)
* Infra: dgit-repos-server; In NOFFCHECK but not FRESHREPO case, do not ↵Ian Jackson2015-05-31
| | | | request wrong ref name
* dgit: Gate sending "previously" on deliberately_not_fast_forward not $forceflagIan Jackson2015-05-31
| | | | | | | $forceflag will only be set if we actually think the ref is not fast forwarding. But the --deliberately flags may cause a repo reset which will require us to mention the previous state, even if the push appears to be ff.
* Push replay prevention protocol: Rename "supersedes" to "previously", as not ↵Ian Jackson2015-05-31
| | | | necessarily actually superseded
* Test suite: add some replay prevention testsIan Jackson2015-05-31
|
* Infra: Document dgit-repos-policy-trusting a bitIan Jackson2015-05-31
|
* Infra: Add dgit-repos-policy-trustingIan Jackson2015-05-31
| | | | | | | | Also add the --deliberately-fresh-repo deliberately, both to the manpage and to dgit's idea of non-fast-forward-ness. Rename the dgit test only deliberatelies to start with TEST-dgit-only- rather than having the -dgit-only at the end.
* Test suite: move t-commit and break out t-policy (nfc)Ian Jackson2015-05-31
|
* Infra: Allow FRESHREPO to override "tag already exists"Ian Jackson2015-05-31
| | | | drs-push-rejects test needs updating too.
* Infra: dgit-repos-server: Fix some typos in debug messagesIan Jackson2015-05-31
|
* Infra: Make replay prevention file contain tags (by searching for right ref ↵Ian Jackson2015-05-31
| | | | names)
* Infra: document DELIBERATELIES parameter to policy hookIan Jackson2015-05-31
|
* Test suite: Provide t-rm-dput-droppingIan Jackson2015-05-31
| | | | | This is needed if tests reuse versions, which we are perhaps going to have to do to test out the replay prevention.
* Test suite: finalise replay preventionIan Jackson2015-05-31
|
* infra: Finish replay preventionIan Jackson2015-05-31
|
* Dgit.pm: git_for_each_ref: Provide optional $git_dir argumentIan Jackson2015-05-31
|
* Infra: New approach to reply prevention hhen NOFFCHECK, involving removed ↵Ian Jackson2015-05-31
| | | | tags file
* dgit: Declare a "supersedes" in the signed tag for all previous refs/heads ↵Ian Jackson2015-05-31
| | | | and refs/tags in the repo, when deliberately_not_fast_forward
* dgit: Break out deliberately_not_fast_forward (nfc)Ian Jackson2015-05-31
|
* dgit-repos-policy-debian: better debugging for tainted object checkingIan Jackson2015-05-31
|
* New approach to replay prevention - WIPIan Jackson2015-05-31
|
* New approach to replay prevention - WIPIan Jackson2015-05-31
|
* Test suite: Abstract t-tstunt-parsechangelogIan Jackson2015-05-31
|
* Test suite: Use stunt Dpkg::Changelog::ParseIan Jackson2015-05-31
| | | | | | This saves a further 13s. Overalll the stunt changelog machinery saves 28s out of 100s (in my example case, the current debpolicy-newreject).
* Test suite: Allow for t-stunt to override Perl modulesIan Jackson2015-05-31
|
* Test suite: Use stunt dpkg-parsechangelogIan Jackson2015-05-31
| | | | | | | | | | This program takes 16ms instead of ~700ms on my computer. This saves 15s out of 100s for the debpolicy-newreject test (in its current state). Currently this is only used by various things in devscripts because dpkg-source uses /usr/lib/dpkg/parsechangelog/debian directly via a Perl module Dpkg::Changelog::Parse.
* Test suite: Provide new t-stunt facilityIan Jackson2015-05-31
|
* policy: Improve taint message recorded in db by --deliberately-not-fast-forwardIan Jackson2015-05-31
|
* git_get_ref: Move to Dgit.pm and reimplementIan Jackson2015-05-31
| | | | | | | | | | | | | Use git for-each-ref --format=%(objectname) [r]efs/whatever This pattern won't match literally as a prefix because a ref name cannot contain [, so it must match according to fnmatch. whatever cannot contain fnmatch metacharacters because they're not valid in ref names. This is a better idiom because it might directly look up the ref in question (although currently it probably doesn't).
* Dgit.pm: git_for_each_ref: do not insist that first component after refs/ ↵Ian Jackson2015-05-31
| | | | matches \w+
* Break out ensuredir and make mkrepotmp use itIan Jackson2015-05-31
|
* Dgit.pm: git_for_each_tag_referring provides referred-to object id tooIan Jackson2015-05-31
|
* Move is_fast_fwd into Dgit.pmIan Jackson2015-05-31
|
* Move git_rev_parse into Dgit.pm (renaming it from rev_parseIan Jackson2015-05-31
|
* Move various useful functions into Dgit.pm. Necessitates some slightly ↵Ian Jackson2015-05-31
| | | | fancy footwork with $::us