manifestmerge: fix order in which manifests are fetched
If the manifest of an earlier revision on the same delta chain is read before
that of a later revision, the revlog remembers that we parsed the earlier
revision and continues applying deltas from there onwards. If manifests are
parsed the other way round, we have to start over from the fulltext.
For a fresh clone of mozilla-central, updating from
29dd80c95b7d to its parent
aab96936a177 requires approximately 400 fewer zlib.decompress calls, which
results in a speedup from 1.10 seconds to 1.05.
Test document extraction
$ "$TESTDIR/hghave" docutils || exit 80
$ HGENCODING=UTF-8
$ export HGENCODING
$ { echo C; find "$TESTDIR/../i18n" -name "*.po" | sort; } | while read PO; do
> LOCALE=`basename "$PO" .po`
> echo
> echo "% extracting documentation from $LOCALE"
> echo ".. -*- coding: utf-8 -*-" > gendoc-$LOCALE.txt
> echo "" >> gendoc-$LOCALE.txt
> LC_ALL=$LOCALE python "$TESTDIR/../doc/gendoc.py" >> gendoc-$LOCALE.txt 2> /dev/null || exit
>
> # We call runrst without adding "--halt warning" to make it report
> # all errors instead of stopping on the first one.
> echo "checking for parse errors"
> python "$TESTDIR/../doc/runrst" html gendoc-$LOCALE.txt /dev/null
> done
% extracting documentation from C
checking for parse errors
% extracting documentation from da
checking for parse errors
% extracting documentation from de
checking for parse errors
% extracting documentation from el
checking for parse errors
% extracting documentation from fr
checking for parse errors
% extracting documentation from it
checking for parse errors
% extracting documentation from ja
checking for parse errors
% extracting documentation from pt_BR
checking for parse errors
% extracting documentation from ro
checking for parse errors
% extracting documentation from ru
checking for parse errors
% extracting documentation from sv
checking for parse errors
% extracting documentation from zh_CN
checking for parse errors
% extracting documentation from zh_TW
checking for parse errors