Re: XEmacs 21.4.10 release candidate 1 is available - PLEASE TEST

Stephen> Please test Release Candidate 1 of XEmacs 21.4.10 Stephen> ================================================= hi stephen i have sent build reports for this release including a motif build incase you needed one i have noticed that some things have speeded up :) notably opening large emails in gnus and it starts up faster i dont have any comparisons or actual stats just observations can i vote for this one as the next stable ? :) i will continue to use this release and let you know if anything breaks or if i notice anything being better i also didnt see the make check failures mentioned below maybe its another one of those FreeBSDisms on the whole happy :) Stephen> Report your builds via M-x build-report and any bugs or interesting Stephen> behavior (including improvements ;-) via M-x report-emacs-bug. Stephen> Besides the main issue (syntax optimizations affecting fontlock, Stephen> indentation, movement, and cperl mode), there are a fair number of Stephen> dialog fixes. If you're not using the most recent Windows version Stephen> (which already has the fixes), you may want to check these out. Stephen> Of special interest: java-mode, cperl-mode, font-lock, indentation Stephen> ------------------------------------------------------------------- Stephen> It is important to test RC1 (this candidate) or Release Candidate 2 Stephen> (there will most likely be an RC2 because there are some important Stephen> patches pending) because it contains much work on the syntax cache Stephen> code, which is responsible for seriously degraded performance. Stephen> This is especially noticable in large (> 2000 line) Java files (on a Stephen> PIII 650MHz class machine, YMMV). Stephen> There are three main issues to watch for. Stephen> o Compare font-lock and indenting performance at the end of large Stephen> Java files when using RC1 against earlier 21.4 releases. Stephen> You should see a large speedup in RC1 with `lookup-syntax-properties' Stephen> set to nil (now the default). There may be speedups in other modes, Stephen> but java-mode seems to be the worst. If performance still seems Stephen> unacceptable, please let us know. Also, if some large files seem to Stephen> have acceptable performance at the end, but others are unacceptable, Stephen> that is important information. The performance problem depends on the Stephen> buffer position; for me, it's only perceptible thousands of lines from Stephen> the beginning of the buffer. Stephen> o Watch for bugs in font-locking, indentation, and syntax-based Stephen> movement (forward-comment, forward-sexp, forward-word, etc). Stephen> To optimize the syntax lookups (ie, `lookup-syntax-properties' Stephen> set to t), a fair amount of surgery was done on src/syntax.c. I have Stephen> done my best to avoid changing semantics, but the code is very Stephen> unclear. Bugs may have been introduced. Stephen> o cperl users must set `lookup-syntax-properties' to t. Stephen> Because this performance problem affects everyone, we have changed the Stephen> default of the Lisp variable `lookup-syntax-properties' to nil. This Stephen> means that modes that depend on looking up `syntax-table' text Stephen> properties will not work properly unless you set the variable to t. Stephen> The only mode I know of that this affects is cperl mode. Stephen> We will be discussing with the cperl maintainers how best to minimize Stephen> inconvenience here, but for RC1, cperl users will have to turn on Stephen> `lookup-syntax-properties' via customize-variable or in their init Stephen> files. Mode maintainers: on Alan Mackenzie's recommendation, we have Stephen> aliased parse-sexp-lookup-properties to lookup-syntax-properties. Stephen> This alias will also be put into the XEmacs 21.5 tree. Stephen> Other issues: Stephen> ------------- Stephen> o Test failures will occur in "make check". Stephen> Please note that in "make check" 6 regexp tests will fail. This is Stephen> due to a variance between the ways "\<" and "\b" are implemented; the Stephen> latter is buggy. This is unrelated to the syntax table changes. It Stephen> will (I expect) be fixed for RC2. -- XEmacs Advocate | Do not try the patience of Wizards, FreeBSD Devote | for they are subtle and quick to anger. Perl Hacker | - Elric (Technomage) , Babylon 5. Apache God | <mailto:rendhalver at xemacs.org> <GnuPG KeyID: AE51D190>