run-tests: only lock Popen wait/poll on python 2.4 It can cause scheduling bubbles and is not necessary on newer pythons.

run-tests: lock popen wait/poll In python2.4, any call to Popen() may attempt to wait on any active process, and wait is not thread-safe. Make it thread-safe. See http://bugs.python.org/issue1731717 for details.

merge with i18n

i18n-de: fixed some strings about stripping

i18n-de: synchronized with fbdac607bff3

rollback: mark as deprecated

worker: properly report errors from worker processes (issue3982)

run-tests: patch over py2.4 proc.wait() race

worker: check problem state correctly (issue3982) If a large update triggered an abort, it was possible for the main thread to still update the dirstate. This fix is incomplete, as the failing worker now doesn't generate a proper error message. This is difficult in the fork-based framework, which relies on exceptions propagating to the top of the dispatcher for formatting.

sshpeer: mark _validrepo internal

progress: respect HGPLAIN

tip: deprecate the tip command

tag: remove incorrect reference to tip

log: remove tip from example

cat: remove incorrect reference to tip

share: remove reference to tip

rebase: simplify references to branch tips

mq: remove reference to tip

hgweb: show current search query in the input field

mq: use an unfiltered property cache for the queue object The same `queue` object should be used in all situations. This allows to store global states on the object (like the reworked shelve extension plans to). The queue object does not need to be distinct for all possible views (no reference to the repo/view used at creation time) it's even a bit surprising that mq still works with the current code.

summary: use missing ancestors algorithm to find new commits For a repository with around 520,000 commits and 190,000 files, this reduces the time hg summary takes from 5.5 seconds to 2.75.

Merge histedit fixes from stable.

histedit: use base for computing renames when folding (issue3729) When a file was renamed in the very first changeset being histedited, the rename was not detected. Consider a history like this: 0 base: add a.txt 1 first: rename a.txt to b.txt 2 last: edit b.txt When 1 and 2 are edited, copies.pathcopies(first, last) correctly returns that nothing was renamed. We must instead use copies.pathcopies(first, last) to detect the initial rename.

test-histedit-fold: remove unnecessary --template The template is already set by the logt alias.

templater: sort functions alphabetically, as filters are

hgweb: color line which is linked to in file source view Browser compatibility (':target' selector used): IE 8+, FF 3.5+, Safari 3.1+, Opera 9.5+, Chromium 4+, all popular mobile browsers.

hgweb: wrap lines in file source view

hgweb: code selection without line numbers in file source view All the source lines are put in a <pre> tag, which gives correct display and copy&paste in both Chromium (WebKit) and FireFox: line numbers are not copied, all the tabs and spaces are kept. This doesn't change the visual appearance of the view compared to current hgweb version and doesn't use any JS code. Also, stripes in this view are now generated clientside with CSS. This implementation is chosen because other variants have important issues: Strategy FF Chrome current D,LT,E,T,L D,L pre S,NW S,NW pre/div/nbsp LT,E,T,TS,NW TS,NW pre/div/br LT,E,T,NW NW ol/li/nbsp LT,E,T,TS,AJ TS,AJ ol/li/br LT,E,T,AJ AJ pre/span LV LV Legend Strategies: - current: implemented in hgweb before this patch, i.e. divs for each line, and line numbers links in the div too - pre: the whole code in one pre tag with newlines, all line numbers in another one with 'float: left' - pre/div/{nbsp,br}: same as just 'pre', but separate divs for each line and &nbsp; or <br> instead of empty lines (otherwise they are not copied at all) - ol/li/{nbsp,br}: a single ol with li's and divs for each line, &nbsp; or <br> same as in previous strategy - pre/span: this patch Problems: D = (very minor) display problems, like wrong width of leading tabs LT = loses leading/trailing whitespace E = loses embedded whitespace B = loses blank lines T = loses tabs L = selects line numbers LV = (only) visually selects line numbers LVE = (only) visually selects line numbers at empty lines S = no stripes (and no ability to easily highlight lines-which-are-linked-at in the future) TS = space copied instead of empty line AJ = get anchor links only with JS (they work even without) NW = no linewrap easily possible (in future) As for browser versions compatibility, the CSS tricks used are supported in (according to caniuse.com): a) line numbers generation with 'content:' property and CSS counters: IE 8+, all other popular browsers (in pre-WebKit Opera numbers are being copied) b) stripes ('nth-child' selector): IE 8+, FF 3.5+, Safari 3.2+, Opera 9.5+, all other popular browsers c) line numbers are not visually selected ('user-select:' property): IE 10+, Opera 15.0+, all other popular browsers This patch is based on a demo implementation by Martin Geisler <martin@geisler.net>.

tests: print line numbers when checking code When an exception will be listed the line numbers can be replaced with glob as it is done in test-check-pyflakes.t

tests: rename files to py extension instead of copying a 2nd time