Thu Sep 18 09:30:17 PDT 2008 Judah Jacobson * Fix hang after a user input error (for example, EOF). New patches: [Fix hang after a user input error (for example, EOF). Judah Jacobson **20080918163017] hunk ./src/Darcs/Utils.lhs 18 formatPath ) where import Prelude hiding ( catch ) -import Control.Exception ( bracket, bracket_, catch, Exception(IOException) ) +import Control.Exception ( bracket, bracket_, catch, Exception(IOException), throwIO ) import Control.Concurrent ( newEmptyMVar, takeMVar, putMVar, forkIO ) #if !defined(WIN32) || __GLASGOW_HASKELL__>=609 import Control.Concurrent ( threadWaitRead ) hunk ./src/Darcs/Utils.lhs 144 (reset_umask rc) job --- withThread is used to allow ctrl-C to work even while we're waiting --- for user input. +-- withThread is used to allow ctrl-C to work even while we're waiting for +-- user input. The job is run in a separate thread, and any exceptions it +-- produces are re-thrown in the parent thread. withThread :: IO a -> IO a withThread j = do m <- newEmptyMVar hunk ./src/Darcs/Utils.lhs 149 - forkIO (j >>= putMVar m) - takeMVar m + forkIO (runJob m) + takeMVar m >>= either throwIO return + where runJob m = (j >>= putMVar m . Right) `catch` (putMVar m . Left) askUser :: String -> IO String #ifdef HAVE_HASKELINE Context: [Disable more record_editor tests under Windows. Eric Kow **20080917192602 It still appears to be failing under Cygwin. ] [add a couple of tests of unrecord. David Roundy **20080917133738] [add a few (passing) tests of pending handling in obliterate and unrecord. David Roundy **20080917130423] [Use putStrLn for "Cancelled." message. Dmitry Kurochkin **20080916205014] [remove now-unused is_addrmfile and is_addrmdir. David Roundy **20080916173136] [resolve issue1012: it seems to be fixed by the fix for issue709. David Roundy **20080916173116] [resolve issue885: fix patchSetToRepository to work with hashed. David Roundy **20080916173030] [resolve issue709: avoid adding changes to pending in rmpend when possible. David Roundy **20080916173002] [first-stage fix for issue709. David Roundy **20080916170333 Here I fix the bug which leads to a corrupt pending being left, with a rmfile change but no hunk removing the contents. This doesn't fix issue709, since an incorrect pending is left, it's just no longer a corrupt pending (i.e. it's still got the rmfile, but if you record it there's no problem). ] [add new test that we don't do anything when repairing a clean repo. David Roundy **20080916165437] [whitespace change in prepend. David Roundy **20080916160425] [make shell_harness print summary of passing tests when running bugs tests. David Roundy **20080916145820] [trailing whitespace cleanup in Repository.Internal. David Roundy **20080916142112] [tiny refactor in Internal.lhs. David Roundy **20080916155922] [simplify issue965 test (which took quite a while for me to figure out). David Roundy **20080916152028] [Test for issue691. Dmitry Kurochkin **20080916111332] [Throw an error when haskeline returns an EOF/EOT. Judah Jacobson **20080917180124] [Re-enable haskeline by default. Require haskeline>=0.3.1 which fixes issue1050. Judah Jacobson **20080917194007] [Make match.pl test understand Windows dates. Eric Kow **20080916011339 Windows does not abbreviate its timezones. ] [Disable some tests that don't work under Windows. Eric Kow **20080916000912] [Translate get.pl test into shell. Eric Kow **20080916000227 The original get.pl uses the Perl Cwd library, which seems not to work for purposes of this test under MSYS. ] [Resolve issue691: distinguish between NoArg and ReqArg in defaults parser. Dmitry Kurochkin **20080915224046] [Move get_default_flag type to definition. Dmitry Kurochkin **20080915220316] [reverse overly-invasive haddocking. David Roundy **20080915110353] [In darcs send if POST fails try sendmail. Dmitry Kurochkin **20080914234314] [Grammar in Darcs.Commands.Send. Dmitry Kurochkin **20080914232923] [Print "Successfully sent patch bundle to" only when we really sent something. Dmitry Kurochkin **20080914232905] [Fix overzealous escaping in haddock. Eric Kow **20080915070926] [Haddock documentation for English.lhs florent.becker@ens-lyon.org**20080913143844 Docstrings by Eric Kow ] [Haddock documentation for PrintPatch florent.becker@ens-lyon.org**20080913143420] [Haddock documentation for Flags florent.becker@ens-lyon.org**20080913143026] [Haddock documentation for Motd florent.becker@ens-lyon.org**20080913141335 docstring by Eric Kow ] [Haddock string for TheCommands florent.becker@ens-lyon.org**20080913141040 Docstring by Eric Kow ] [Run testing in temp directories to avoid collosions, fail if DarcsURL header is not found. Dmitry Kurochkin **20080914223930] [Use tempfile() UNLINK option to automatically remove temp files at exit. Dmitry Kurochkin **20080914223827] [Coding style in upload.cgi. Dmitry Kurochkin **20080914223751] [Stop after we found the first DarcsURL: in patch bundle. Dmitry Kurochkin **20080914220421] [Spaces in upload.cgi. Dmitry Kurochkin **20080914220324] [Really allow pull.sh test to pass on Windows. Eric Kow **20080914211458 It seems that these redirections cause an openBinaryFile failure under Windows. I do not yet understand why using different names makes a difference, but it could provide an interesting clue. ] [rewrite upload.cgi so it won't need any customization by default. David Roundy **20080913171447 The downside is that it has to do a darcs get --lazy in order to check if the patch can be applied. If you define $target_repo, however, it doesn't need to do this (but then can only accept patches to a single locally-present repo). ] [when _darcs/prefs/post is present, use _darcs/prefs/post for To: header. David Roundy **20080913171025] [sketchy documentation of _darcs/prefs/post David Roundy **20080913115655] [ChangeLog entries for darcs 2.0.3pre1 Eric Kow **20080914094144] [Disable amend-cancelling test under Cygwin. Eric Kow **20080913213039] [Make binary.sh test more portable (avoid copying an exe). Eric Kow **20080913212843 Under Windows, copying the exe file will result in an .exe extension being appended to the filename. This confuses the test. ] [set default upload.cgi to work on darcs-unstable. David Roundy **20080913112227] [Improve upload.cgi. Dmitry Kurochkin **20080912230953 - check if patch is valid before sending - use sendmail to send patches or drop to maildir ] [Spaces and punctuation in upload.cgi. Dmitry Kurochkin **20080912230827] [fix problem in Darcs.URL refactor pointed out by Eric. David Roundy **20080913104327] [Disable a pull.sh test under Windows. Eric Kow **20080912224027 It relies on darcs not working if we chmod u-r a file. This seems to have little effect in Windows. ] [refactor Darcs.URL to eliminate use of Regexes. David Roundy **20080912173611 The algorithms here are not tricky, and I find this easier to read. ] [change is_file to return false on droundy@darcs.net: David Roundy **20080912173501] [clean up whitespace. David Roundy **20080912150248] [fix manual for optional arguments. David Roundy **20080912150231] [clean up whitespace. David Roundy **20080912145708] [add test for new --output-auto-name feature. David Roundy **20080912145648] [Spaces in Darcs.Commands.Send module. Dmitry Kurochkin **20080912122223] [Make '--output-auto-name' accept optional directory argument. Dmitry Kurochkin **20080912120516] [Add DarcsOptAbsPathOption for options with optional path argument. Dmitry Kurochkin **20080912112458] [Refactor Darcs.Repository.Prefs.getCaches. Dmitry Kurochkin **20080912055204] [Print warning when '--http-pipelining' option is used, but darcs is compiled without HTTP pipelining support. Dmitry Kurochkin **20080912054253] [Do not download URL we have speculated before. Dmitry Kurochkin **20080912053236] [Spaces and parentheses in URL module. Dmitry Kurochkin **20080912053000] [Coding style in Darcs.Arguments.network_options. Dmitry Kurochkin **20080911140710] [Resolve issue1054: --no-cache option to ignore patch caches. Dmitry Kurochkin **20080911140233] [Remove unused variable from configure.ac. Dmitry Kurochkin **20080911132107] [Comment in configure.ac. Dmitry Kurochkin **20080911115840] [Indentation fixes in configure.ac. Dmitry Kurochkin **20080911115117] [insecure printfs - fix the two that I can currently hit Steve Cotton **20080910230659] [Formating and minor refactoring in URL.urlThread. Dmitry Kurochkin **20080910061227] [TAG this version works. David Roundy **20080910212908] Patch bundle hash: 0c87acd42b34b96bc49c61a8eeea6095853636c8