[Author Prev][Author Next][Thread Prev][Thread Next][Author Index][Thread Index]

gEDA-cvs: pcb.git: branch: pcb-20110918 created (2a05dd78fb05c967845933591ad21183c259445a)



The branch, pcb-20110918 has been created
        at  2a05dd78fb05c967845933591ad21183c259445a (commit)


commit 2a05dd78fb05c967845933591ad21183c259445a
Author: DJ Delorie <dj@xxxxxxxxxxx>
Date:   Sun Sep 18 15:13:18 2011 -0400

    Today is the 18th :-)

commit e7881e1cd1ddffde9d685c32661e713caf715c42
Author: DJ Delorie <dj@xxxxxxxxxxx>
Date:   Sun Sep 18 15:10:58 2011 -0400

    Update Changelog and NEWS for 20110918

commit 9b21e7d14d2d5e86c7ffd76972e362bc1d2ebdb0
Author: DJ Delorie <dj@xxxxxxxxxxx>
Date:   Sun Sep 18 13:38:01 2011 -0400

    Remove src/hid/common/hidgl_package_vrml_y.c from POTFILES.in
    
    Reverts 61bd0cb1e134bbea96abd6fd2d5fb94d09c0b3c8 as this file doesn't
    exist in this repository, causing a distcheck failure.

commit df25b6b3f1326a2adccfbf1e20223c4cfa8a4057
Author: DJ Delorie <dj@xxxxxxxxxxx>
Date:   Sun Sep 18 13:20:55 2011 -0400

    Update golden files for nanometers
    
    hid_png3 was off by a pixel on a line end.  I retested my UV and TT
    films to verify that the nanometers conversion didn't affect line
    *widths* and updated the golden file to accept this one-pixel
    difference.

commit c8214a0ca40ed31ea31a98904c798dc8f7bfe322
Author: DJ Delorie <dj@xxxxxxxxxxx>
Date:   Sat Sep 17 17:28:33 2011 -0400

    [windows] use unix-style path separators when calling gnetlist

commit c9b8eaee755a5e192732814e275f6a0ce67386b3
Merge: 70db91fc3dfb8b378962794fc8f86f71b52f7af8 7452c9e8fc227353e9fec7b2a8d0b5014a9ba828
Author: Felix Ruoff <Felix@xxxxxxxxxxxxxxxxxx>
Date:   Sat Sep 17 22:50:17 2011 +0200

    Merge branch 'master' of ssh://git.gpleda.org/pcb

commit 70db91fc3dfb8b378962794fc8f86f71b52f7af8
Author: Kai-Martin Knaak <kmk@xxxxxxxxxxxx>
Date:   Fri Sep 16 01:07:32 2011 +0200

    remove panner control from documentation
    
    There is no panner control in the GUI (anymore?) since at least five years.
    this pathc removes its description from the texi manual.
    
    Closes-bug: lp-851443

commit 7452c9e8fc227353e9fec7b2a8d0b5014a9ba828
Author: Kai-Martin Knaak <kmk@xxxxxxxxxxxx>
Date:   Fri Sep 16 01:07:32 2011 +0200

    remove panner control from documentation
    
    There is no panner control in the GUI (anymore?) since at least five years.
    this pathc removes its description from the texi manual.

commit 07d0e2691bc9dbbf5eb743468cce5ab82f28f32a
Author: Peter Clifton <pcjc2@xxxxxxxxx>
Date:   Sat Sep 17 16:08:59 2011 +0100

    Don't adjust the Library command settings if they are NULL, or
    an empty string.
    
    (The empty string part is a fixup for
    commit 31df66e87a810aac852139da9119b15c47fa8c71
     - this will teach me for bikeshedding)
    
    The NULL part should fix bug lp-852598
    
    Closes-bug: lp-852598

commit 31df66e87a810aac852139da9119b15c47fa8c71
Author: Peter Clifton <pcjc2@xxxxxxxxx>
Date:   Sat Sep 17 15:13:09 2011 +0100

    Bikeshed commit 11700ba8c3c951a788c8190073eb09822060dfdd
    
    I don't like magic strings, and IMO, using a blank LibraryContentsCommand
    string to imply "don't use this feature" is better than "*NONE*".

commit aa73c06889aa6c8ac7b7ff5816ee4863915c7994
Author: Peter Clifton <pcjc2@xxxxxxxxx>
Date:   Sat Sep 17 14:52:41 2011 +0100

    hid/gtk: Set the LC_NUMERIC locale to "C" rather than "POSIX"
    
    This value is supported on Win32 platforms, whereas "POSIX" doesn't
    appear to have any effect.
    
    This was one possible way to fix the "zoom doesn't work" bug on Win32
    for locales which use "," as the decimal separator.
    
    DJ fixed the Zoom action to be locale independent in
    commit fd5399c67b988f2f7c9d1a0b4ab7c13bc2e95158, which worked around
    the problem for this particular action, but I feel it would be best
    that we made our locale handling consistent between Linux and Win32
    as well.
    
    Affects-bug: lp-843577

commit 09aab0bb01e914cc773d2115292695645f561bd5
Author: Dima Kogan <dima@xxxxxxxxxxxxxxx>
Date:   Thu Sep 8 14:31:13 2011 -0700

    fixed an instance of the "only-optimize-autorouted" flag being ignored
    
    Signed-off-by: Dima Kogan <dima@xxxxxxxxxxxxxxx>

commit 11700ba8c3c951a788c8190073eb09822060dfdd
Author: DJ Delorie <dj@xxxxxxxxxxx>
Date:   Fri Sep 16 23:30:41 2011 -0400

    Skip the M4 library on Windows
    
    On Windows, set the LibraryContentsCommand to the magic string "*NONE*"
    and check for that string while loading the libraries.  This avoids the
    windows pcb.exe from trying to run bash or m4 scripts, which are
    normally not available, yet allows the user a way to re-enable them if
    they want.

commit fd5399c67b988f2f7c9d1a0b4ab7c13bc2e95158
Author: DJ Delorie <dj@xxxxxxxxxxx>
Date:   Fri Sep 16 14:08:04 2011 -0400

    De-localize zoom parsing
    
    The values used for Zoom() were being parsed with strtod(), which
    switches between '.' and ',' depending on locale.  Switch to
    g_ascii_strtod() which always uses '.' and thus always matches our
    actions.
    
    Note: I didn't change the command line parsing, because I didn't think
    we should impose '.'-centric syntax on other locales.

commit a2bcea72cf4161eb928582a28fac3443309d8050
Author: DJ Delorie <dj@xxxxxxxxxxx>
Date:   Fri Sep 16 01:23:25 2011 -0400

    Add DIL footprint names and R025.
    
    Existing tutorials say to use "DIL 16 300" for DIPs (we use "N 16 300")
    and often call for the R025 footprint, so add them as aliases so the
    m4-to-newlib converter includes them.

commit 8147be7a70654c6c29d0e29e222c4014cf19fa5d
Author: DJ Delorie <dj@xxxxxxxxxxx>
Date:   Fri Sep 16 01:01:00 2011 -0400

    Fix m4-to-newlib converter filenames
    
    The m4 library uses macros and parameters to describe footprints, which
    allows for a *very* flexible footprint naming scheme.  While it is
    impossible for pcb to know all valid footprint values, it does have a
    list of "standard" ones, and knows how to obtain those.  However, the m4
    to newlib converter wasn't using the same footprint naming convention as
    everyone else, so produced *.fp files that weren't always useful.  For
    example, there were a large number of 300.fp files that all collided.
    
    With this fix, the footprint name matches the value in [...] in the
    library window, which is what we tell people to use for the "footprint="
    attribute anyway.

commit a8b69632f0219923fdbc9df8c44abeed7356b495
Author: DJ Delorie <dj@xxxxxxxxxxx>
Date:   Fri Sep 16 01:00:32 2011 -0400

    Wrap all XmStrings in gettext()

commit 61bd0cb1e134bbea96abd6fd2d5fb94d09c0b3c8
Author: Peter Clifton <pcjc2@xxxxxxxxx>
Date:   Thu Sep 15 23:32:33 2011 +0100

    Add another missing source file to POTFILES.in

commit fa0768e9a8a1367fedcac74858fe3d08a89f8a71
Author: DJ Delorie <dj@xxxxxxxxxxx>
Date:   Thu Sep 15 15:23:27 2011 -0400

    Add locale functions
    
    Add textdomain() and setlocale() to connect to our message catalogs.

commit 8b130d1ee31a41df5dc90c0a729e902b33e11976
Author: Peter Clifton <pcjc2@xxxxxxxxx>
Date:   Thu Sep 15 18:57:32 2011 +0100

    Add some missing files to POTFILES.in
    
    (And sorted the file contents).
    
    Reported-by: Bert Timmerman <bert.timmerman@xxxxxxxxx>
    Affects-bug: lp-846368

commit dd644140ce265bb87a71ff0855a9b1a7083047cb
Author: Bert Timmerman <bert.timmerman@xxxxxxxxx>
Date:   Fri Sep 9 23:12:23 2011 +0200

    Updated the Dutch translation.

commit a537b78f5edc8c8c106ee146987797ef99ceb60a
Author: Peter Clifton <pcjc2@xxxxxxxxx>
Date:   Thu Sep 15 00:24:17 2011 +0100

    report.c: Don't leave actions in the undo stack or change the user's flags
    
    Some actions in ReportNetLength*() would have been lumped in with
    whatever actions were next added to the undo stack. These functions
    also left the user's board with a different "FOUND" flags on various
    items.
    
    In the case of the "Report(AllNetLenths)" action, a very long undo stack
    would be created, which is not very helpful to the user.
    
    These changes hide the internal flag changes made during the net length
    calculation from the undo system.
    
    We are able to get back to the inital board state because we save the
    operations caused by our initial RestConnections() on the undo stack.
    After we have done our work (modifying various flags in the process),
    we reset the flags again (this time bypassing the Undo system), so the
    board is consistent with the ResetConnections() call we DID save undo
    data for. We then Undo() these changes before we leave the function.

commit 9a8e559401db73ea3c9f9ecddb9f50a179a91bb2
Author: Peter Clifton <pcjc2@xxxxxxxxx>
Date:   Thu Sep 15 17:06:15 2011 +0100

    report.c: Return 1 for all failure conditions in ReportNetLengthByName()

commit d0a3105c8dfe80d6a03d358ae850a9587236a23f
Author: Peter Clifton <pcjc2@xxxxxxxxx>
Date:   Wed Sep 14 23:27:45 2011 +0100

    undo.c: Add error message where RestoreUndoSerialNumber() will break things
    
    If operations are addded to the undo stack between a call to
    IncrementUndoSerialNumber() and a subsequent call for
    RestoreUndoSerialNumber(), those added operations will be placed in the
    undo stack with an inconsistent serial number.
    
    The {Save,Restore}UndoSerialNumber() API is pretty fragile in this
    regard, and we should avoid using it where possible.
    
    An better alternative might be to implement a "LockUndoSerialNumber()"
    call which temporarily disables IncrementUndoSerialNumber(), and a
    "UnlockUndoSerialNumber()" call which re-enables its increment function.
    
    Better yet - we could make these functions nestable, so we need not
    worry whether some action uses this new feature internally when we are
    also using it.

commit a7d774b4e0fc7c2aa86a12819c169bc1392b20de
Author: Peter Clifton <pcjc2@xxxxxxxxx>
Date:   Wed Sep 14 21:58:28 2011 +0100

    undo.c: Handle undo failures in a more sane manner
    
    Our current error-case handling can cause serial number inconsistency
    when something goes wrong when performing an Undo operation. There are
    also various logic flaws in our handling, as it only catches failures
    where every sub-undo operation with that serial number fails.
    
    Remove the confusing do-loop which decrements the serial number in the
    failure case, and return some sensible error message instead.
    
    This patch also removes the code which looked like it was intended to
    loop over the undo stack until a serial number was found to operate on.
    It is advantageous for code like the DRC to be able to add increment
    the undo serial number, then safely perform an Undo - even if there were
    no changes logged in the undo system during the previous serial number.
    
    Closes-bug: lp-848509

commit 7705118c47da4941c55d644e20b937e981168e7b
Author: Peter Clifton <pcjc2@xxxxxxxxx>
Date:   Wed Sep 14 21:49:31 2011 +0100

    undo.c: Allow undo of locked objects
    
    Our internal actions can make changes to locked objects, then
    use the undo mechanism to revert these changes. An example is
    DRC, which changes flags on objects and uses the Undo system
    to restore their original values.
    
    Affects-bug: lp-848509

commit d63f3f35cb2d0f6d0bef71ecb4bf17208ed76ad7
Author: Peter Clifton <pcjc2@xxxxxxxxx>
Date:   Wed Sep 14 23:02:02 2011 +0100

    find.c: Remove stray RestoreUndoSerialNumber() calls
    
    These will cause havoc with the undo system, as we don't actually
    save a serial number to restore to.
    
    Until a commit efd212c1deb264e9a7f2be17e9338fbb60e22cc0 we were
    saving a serial number at the start of each "ResetConnections (true);"
    call, and it would have been that serial number which got restored.
    
    With this and some other fixes to the undo system, these restores
    are no longer required.
    
    Affects-bug: lp-848509

commit c8e0f875155405bdcdb6835e8439a2c2fd4fb4a0
Author: Peter Clifton <pcjc2@xxxxxxxxx>
Date:   Wed Sep 14 21:40:27 2011 +0100

    gtk: Fix ghid_zoom_view_fit() to work when the board is flipped.
    
    Affects-bug: lp-850311

commit 0ac0076bc9af38dd07700cbb7a6d3f21a8fc2621
Author: Peter Clifton <pcjc2@xxxxxxxxx>
Date:   Tue Sep 13 19:15:02 2011 +0100

    Fixup OpenGL includes for better compatibility with OSX
    
    Not tested ;)

commit 90ed0e3f7f467cacd7d858daee023b5aa5e417b9
Author: Peter Clifton <pcjc2@xxxxxxxxx>
Date:   Tue Sep 13 19:15:02 2011 +0100

    Better autoconf tests for GL and GLU (from GtkGLExt - GPL v2 or later)
    
    These tests should give better cross platform support for our build.

commit a1197b36497057088057dd0dc4135d19fefda116
Author: Peter Clifton <pcjc2@xxxxxxxxx>
Date:   Tue Sep 13 19:15:02 2011 +0100

    Remove the m4 directory from the root .gitignore file and add its own

commit 1ffdc656f5c601db9b4625b9aad14b4d33a49868
Author: DJ Delorie <dj@xxxxxxxxxxx>
Date:   Tue Sep 13 01:55:39 2011 -0400

    [lesstif] set units for styles dialog in more places
    
    The styles dialog was segfaulting as the units hadn't been set yet, if
    you open the style dialog before doing anything with the grid.

commit 6bbcd9e693a26816abccbd8acdb95100d23c01f9
Author: Peter Clifton <pcjc2@xxxxxxxxx>
Date:   Tue Sep 13 01:21:47 2011 +0100

    hid/gtk: Remove misplaced (and redundant) hidgl_flush_triangles() call.
    
    If we were to flush the triangle buffer in ghid_show_crosshair(), we
    should do it before we set XOR mode on the GL context (assuming the
    buffer might be dirty when we are called), and we should do it after
    we finish drawing (if we wish to leave the buffer clean when we return).
    
    The location the flush was in before this commit is just plain wrong.
    
    As we currently have flushes of the triangle buffer wrapping the call
    to ghid_show_crosshair(), just leave the flushing out of this function.
    
    Does this commit message provide an adequate example of simple patch,
    complex commit message phenomenon?

commit ff9abb9212b375ee7affdffe17803039a8525cfc
Author: Peter Clifton <pcjc2@xxxxxxxxx>
Date:   Tue Sep 13 01:01:45 2011 +0100

    hid/gtk: Remove auto-pan feature (kill it with fire).
    
    This seems to be fairly universally unpopular on geda-user, and its
    implementation is a little fragile, often leading the board to pan off
    to some corner of the board as the poor user reaches for the layer
    selector or a menu without having remembered to click the feature off.
    
    The fact that the pan direction and speed was set the instant you leave
    the drawing window also made it particularly unhelpful. A proper
    implementation should have grabbed the mouse - or at least implemented
    some buffer zone in which the mouse could be used to give feedback on
    the pan direction.

commit be3aec15d207ca92ee23553c93cce957751343eb
Author: DJ Delorie <dj@xxxxxxxxxxx>
Date:   Sun Sep 11 22:10:45 2011 -0400

    [windows] set GDK_PIXBUF_MODULE_FILE
    
    If we're using a local gdk-pixbuf installation, we need to set this env
    variable to point to it.

commit 732e10a394d7d3210f7b70dcfafc08b6c40c5210
Author: Bdale Garbee <bdale@xxxxxxx>
Date:   Sun Sep 11 21:06:13 2011 -0400

    Invoke eps2png in $(srcdir), no .

commit cb95aa44bfde8231ab80ae3642a2f380636b4f8d
Author: Peter Clifton <pcjc2@xxxxxxxxx>
Date:   Mon Sep 12 01:49:47 2011 +0100

    hid/common: Remove #include "draw.h" from hidgl.c (twice!)

commit 2aeb72bfcee45672651a9bcf2950785e67b578a6
Author: Dima Kogan <dima@xxxxxxxxxxxxxxx>
Date:   Sun Sep 11 14:54:00 2011 +0200

    Re-ordered all references in the style selector dialog to have one consistent ordering.
    
    Everything that refers to the settings in the style selector dialog now does so
    in order of
    
    Line thickness
    Hole size
    Hole thickness (hole + annular ring)
    keepaway region size
    
    This wasn't 100% consistent previously, which resulted in a bug (fixed in a
    previous patch). This patch doesn't touch the functionality, but improves the
    style to avoid future bugs.
    
    Modified to apply on recent route style selector changes by Bert Timmerman.
    
    Signed-off-by: Andrew Poelstra (local) <apoelstra@xxxxxxxxxxxxxx>
    Affects-bug: lp-844635

commit 59ee4898847f16cec5e08a40229c5108290fc1f8
Author: Dima Kogan <dima@xxxxxxxxxxxxxxx>
Date:   Sun Sep 11 08:34:59 2011 -0700

    Swap via-hole and -ring size connection in route style selector dialog
    
    Closes-bug: lp-844635
    
    Signed-off-by: Andrew Poelstra (local) <apoelstra@xxxxxxxxxxxxxx>

commit c62863b2b6e8da2c60c9a7d0d5464582563f7b58
Author: Andrew Poelstra (local) <apoelstra@xxxxxxxxxxxxxx>
Date:   Sat Sep 10 23:41:57 2011 -0700

    Repopulate route style selector on PCBChanged action
    
    When loading a new pcb, Bad Things happen to the route style
    selector because pcb's route style data is freed. This patch
    removes all styles from the route style selector and re-adds
    them whenever a new pcb is loaded.

commit 8876712e765a81629d8437664565fedd0945ecb8
Author: Andrew Poelstra (local) <apoelstra@xxxxxxxxxxxxxx>
Date:   Sat Sep 10 23:41:10 2011 -0700

    Add ghid_route_style_selector_empty to GHidRouteStyleSelector

commit 499fe7a6b88a526f6b48d671107ac3d9ec725290
Author: Andrew Poelstra (local) <apoelstra@xxxxxxxxxxxxxx>
Date:   Sat Sep 10 23:28:03 2011 -0700

    Add finalize function to GHidRouteStyleSelector

commit 9dde48253c590a80a18d4a0f12e3f90c1cfedb26
Author: Bert Timmerman <bert.timmerman@xxxxxxxxx>
Date:   Fri Sep 9 18:58:18 2011 +0200

    Remove the src/hid/gtk/gui-dialog-size.c entry from the translatable files list.
    
    This is a completion of commit:
    a914d5ba58104d8559395b66798b35761afb991a
    "Replace old route style selector with GHidRouteStyleSelector"
    dated 20110903 11:09 PM

commit 7d7b21bc0a770553fdf6cd70bb3eeab4c06f7071
Author: Peter Clifton <pcjc2@xxxxxxxxx>
Date:   Fri Sep 9 01:59:06 2011 +0100

    hid/gtk: Filter out double clicks on the layer selector widget
    
    This makes things behave a little nicer if you click to toggle a layer's
    visibility twice in quick succession. Previously, this would be
    interpreted as a double-click, causing an extra synthetic click event
    which would toggle the visibility an extra time.

commit a3b303c83ceb535e6068623b77eb10b14c0ca97a
Author: Peter Clifton <pcjc2@xxxxxxxxx>
Date:   Wed Sep 7 21:23:50 2011 +0100

    hid/gtk: Update file-changed notification bar to match gedit a bit more
    
      1. Add a stock "refresh" image to the "Reload" button.
      2. Change the message text when the board has modifications.
      3. Reload without further prompting if the user presses "Reload",
         even if the board is modified. (We already warned them due to 2.)

commit ba4c4e33a07491a0b1952d600f4166c89f4b54ab
Author: Peter Clifton <pcjc2@xxxxxxxxx>
Date:   Wed Sep 7 21:23:50 2011 +0100

    hid/gtk: Reimplement file-change monitor without GFileMonitor
    
    I must have had a brain-fail when I hooked up setting up the file-changed
    monitor in the function which sets the window title.
    NB: That also gets called after every menu operation!
    
    A better place to hook up the monitor would be ghid_sync_with_new_layout(),
    however changing this alone revealed another issue - we would get notified
    for changes WE make to the files. We were avoiding those events as the
    file-monitor was being reset before it could pop up, at the end of the
    menu action which invoked the save).
    
    However - due to a race condition bug in GLib / GIO, we would sometimes
    see change notify events for changes we made, even when we hooked up
    the GFileMonitor AFTER having saved (and synced) our changes to disk.
    
    Rather than attempt to work around this bug, implement the file-change
    notification in a much more simple way - look at the file mtime when we
    update the layout, and at each time the mouse pointer enters the main
    window. FWIW, this is how gedit currently achieves the same feature.

commit 6762885654463db6b39e6a6dfbc1911d196146fc
Author: DJ Delorie <dj@xxxxxxxxxxx>
Date:   Tue Sep 6 23:39:17 2011 -0400

    Warn user if footprints can't be found on import
    
    Since the message log doesn't make it obvious, add a pop-up dialog
    whenever there are missing footprints during an import.
    
    Affects-bug: lp-828388

commit d46ff633fa3f700c3314a4043dc60fa40b4d827f
Author: DJ Delorie <dj@xxxxxxxxxxx>
Date:   Tue Sep 6 01:56:33 2011 -0400

    Fix math error in IsPointOnLine()
    
    The math in C didn't match the math in the comment (and, apparently,
    hasn't for quite some time).
    
    To test: Draw three lines, from 1000,3000 to 1500,2500, then to
    2000,3000, then back to the start at 1000,3000.  Without this patch, the
    first and last lines are removed and replaced with a copy of the second
    line.

commit 23fdbee2cd45accc1f11856e47508cc49251687e
Author: Peter Clifton <pcjc2@xxxxxxxxx>
Date:   Mon Sep 5 22:55:37 2011 +0100

    Implement a new RevertPCB() call, sharing code with LoadPCB()
    
    The RevertPCB() takes no filename parameter, and aside from that,
    the only difference to LoadPCB is that when it calls the GUI's
    "PCBChanged" action, it passes a new "revert" argument. This should
    let the GUIs optionally do less work resetting the view state for a
    board being reverted.

commit 84ae26807325c8d2d95cd3ec3c2a6e61c217626e
Author: Peter Clifton <pcjc2@xxxxxxxxx>
Date:   Tue Sep 6 00:22:51 2011 +0100

    Add a HID API call, notify_pcb_filename_change()
    
    This is so the core can tell the GUI when the PCB being edited is saved
    into a different file.





_______________________________________________
geda-cvs mailing list
geda-cvs@xxxxxxxxxxxxxx
http://www.seul.org/cgi-bin/mailman/listinfo/geda-cvs