Community
Participate
Working Groups
Created attachment 266477 [details] screen shot of refs spec dialog with no icons. I suspect the is fundamentally an SWT problem ... but, as far as I know is related to the types of icons you use, or something. I will attach a screen shot showing the "force" and "remove" icons are missing from the refs spec dialog screen. Luckily, the function still works if I click in that general area, but, it is kind of hard to know to do that if I was just starting with this "invisible" version. I am running a base system of "Neon.2" which has EGit 4.4.1. BTW, in the "console" there are messages that say *** BUG *** In pixman_region32_init_rect: Invalid rectangle passed Set a breakpoint on '_pixman_log_error' to debug And, that is literally the console, which I always start and leave open, with -console. There is nothing in the metadata/.log file.
David, I see exact same dialog on Windows. "remove" button has an icon, "force" has no icon, but you mean, both are missing now?
(In reply to Andrey Loskutov from comment #1) > David, I see exact same dialog on Windows. > > "remove" button has an icon, "force" has no icon, but you mean, both are > missing now? Yes. I looked on another machine, and see I misspoke, a little, It is one icon and one checkbox that are missing, not two icons.
Created attachment 266503 [details] ref spec dialog on Ubuntu 12.04 showing icon and checkbox This screen shot shows what I was used to seeing, on my Ubuntu 12.04 box. I had recently moved up to 16.04, on another box, and this was one difference that jumped out at me. HTH
I can confirm that on Windows with latest SWT/EGit the buttons are there. David, can you please update to latest egit (4.6 or nightly) and check that the buttons are missing? It looks like GTK issue, so I've moved the bug to SWT.
(In reply to Andrey Loskutov from comment #4) > I can confirm that on Windows with latest SWT/EGit the buttons are there. > David, can you please update to latest egit (4.6 or nightly) and check that > the buttons are missing? It looks like GTK issue, so I've moved the bug to > SWT. Confirmed still "missing" (check box nor icon visible, for this one dialog). That's running the latest I-build, with latest EGit from http://download.eclipse.org/egit/updates/ (which is EGit 4.6.0). Here is what Ubuntu (Linux) says about the version of GTK I am running: $ dpkg -l libgtk2.0-0 libgtk-3-0 Desired=Unknown/Install/Remove/Purge/Hold | Status=Not/Inst/Conf-files/Unpacked/halF-conf/Half-inst/trig-aWait/Trig-pend |/ Err?=(none)/Reinst-required (Status,Err: uppercase=bad) ||/ Name Version Architecture Description +++-===========================-==================-==================-=========================================================== ii libgtk-3-0:amd64 3.18.9-1ubuntu3.1 amd64 GTK+ graphical user interface library ii libgtk2.0-0:amd64 2.24.30-1ubuntu1.1 amd64 GTK+ graphical user interface library ii libgtk2.0-0:i386 2.24.30-1ubuntu1.1 i386 GTK+ graphical user interface library I did double check, and I do not have any SWT_GTK3 variable defined. And, yes, I am running 64 bit. :) Will attach "properties".
Created attachment 266504 [details] properties as copied from page in about dialogs "inst details" properties ... in case it helps. (Since when did they get so large? Maybe too many "updates"?) FYI in this Oxygen version, I also see this "gtk bug" in my *console* window, but does not seem to show up exactly when I pop up that dialog? And nothing in log. *** BUG *** In pixman_region32_init_rect: Invalid rectangle passed Set a breakpoint on '_pixman_log_error' to debug Is the "region32" reason to be suspicious of the "bitness" of libraries which get attached?
(In reply to David Williams from comment #5) > > Here is what Ubuntu (Linux) says about the version of GTK I am running: > Here is what the 12.04 machine says: $ dpkg -l libgtk2.0-0 libgtk-3-0 Desired=Unknown/Install/Remove/Purge/Hold | Status=Not/Inst/Conf-files/Unpacked/halF-conf/Half-inst/trig-aWait/Trig-pend |/ Err?=(none)/Reinst-required (Status,Err: uppercase=bad) ||/ Name Version Architecture Description +++-===========================-==================-==================-=========================================================== ii libgtk-3-0:amd64 3.18.9-1ubuntu3.1 amd64 GTK+ graphical user interface library ii libgtk2.0-0:amd64 2.24.30-1ubuntu1.1 amd64 GTK+ graphical user interface library And since the "region32" message sounded suspicious I removed the i386 gtk packages from by 16.04 install -- not sure what needed them, if anything. But no difference -- in the checkbox and icon displaying. I no longer saw the *** BUG *** In pixman_region32 message in console, though. So, maybe an unrelated issue. But after removing gtk2-engines-pixbug:i386 libgtk2.0-0:i386 gtk2-engines-murrine:i386 Here is what my 16.04 machine says it has: $ dpkg -l libgtk2.0-0 libgtk-3-0 Desired=Unknown/Install/Remove/Purge/Hold | Status=Not/Inst/Conf-files/Unpacked/halF-conf/Half-inst/trig-aWait/Trig-pend |/ Err?=(none)/Reinst-required (Status,Err: uppercase=bad) ||/ Name Version Architecture Description +++-===========================-==================-==================-=========================================================== ii libgtk-3-0:amd64 3.18.9-1ubuntu3.1 amd64 GTK+ graphical user interface library ii libgtk2.0-0:amd64 2.24.30-1ubuntu1.1 amd64 GTK+ graphical user interface library As I said, no difference if visibility of those items. = = = = = = = = = I should probably also mention, another difference between machines is that the 16.04 machine is using a 2560x1440 resolution screen, which is says has a dpi of 96. I was using it for a while with 12.04, and think I would have noticed those icons missing, though. But my remaining 12.04 machine is 1920x1080. I think unrelated, though, since if I switch the 16.04 machine to a lower resolution, I continue to have the same issue. = = = = = = = Sorry for all the data that leads nowhere. Just trying to help rule out issues or narrow down the source.
One more bit of data that leads nowhere: just to be sure, I install a fresh SDK M5, with just EGit added (latest from 'oxygen repo') and still no joy.
David, this should be fixed in latest I-build. Would you please verify?
I can reproduce this issue on SWT master as of today, with GTK3.22 and Fedora 27. I have a JFace snippet that can reproduce the issue: simply run Snippet051TableCenteredImage. On GTK3 the images aren't drawn at all, on GTK2 they are. I'll continue to investigate.
Looks like another elusive drawing bug, GTK3.10+ only. GTK3.8 seems to work fine.
After some investigating, I have discovered that this bug is caused by a regression from bug 438505. This specifically is what breaks it: if (parent.getHeaderVisible() && GTK.GTK_VERSION > OS.VERSION(3, 9, 0)) { r.y += parent.getHeaderHeightInPixels(); } I'll continue to investigate a fix that doesn't re-introduce bug 438505.
(In reply to Eric Williams from comment #12) > After some investigating, I have discovered that this bug is caused by a > regression from bug 438505. This specifically is what breaks it: > > if (parent.getHeaderVisible() && GTK.GTK_VERSION > OS.VERSION(3, 9, 0)) { > r.y += parent.getHeaderHeightInPixels(); > } > > I'll continue to investigate a fix that doesn't re-introduce bug 438505. The bug seen in EGit/Snippet051TableCenteredImage stems from a coordinate issue. The fix for bug 438505 introduces logic to adjust to paint on the SwtFixed's GdkWindow instead of the GtkTreeView's GdkWindow. Since this window includes any enabled Table headers, the Table geometry API (getClientArea(), getItem(Point p), etc.) was patched to adjust for the height of any Table headers. This leads to the bug: because EGit/Snippet051 aren't looking up the TableItem via coordinates (Table.getItem(Point p)), the drawing area bounds are being adjusted for the height of the header when no such adjustments are needed. This is a bug in SWT because it breaks the TableItem.getBounds() API. I've reverted the fix for bug 438505 locally and I'm currently bisecting GTK to find the issue that broke it originally.
New Gerrit change created: https://git.eclipse.org/r/117300
Gerrit change https://git.eclipse.org/r/117300 was merged to [master]. Commit: http://git.eclipse.org/c/platform/eclipse.platform.swt.git/commit/?id=8a745bffa40230e73fb229950d6d0520b474c8f3
(In reply to Eclipse Genie from comment #15) > Gerrit change https://git.eclipse.org/r/117300 was merged to [master]. > Commit: > http://git.eclipse.org/c/platform/eclipse.platform.swt.git/commit/ > ?id=8a745bffa40230e73fb229950d6d0520b474c8f3 Fix is in master now.
Thanks for bugfix.
*** Bug 493213 has been marked as a duplicate of this bug. ***
*** Bug 529563 has been marked as a duplicate of this bug. ***
We've just stumbled upon the table editing regression on GTK3 in 4.7.2 and I've bissected SWT to this bug which *fixes* the problem for us. Eric, I guess this will be too late for a call for backporting this patch to 4.7.3?
(In reply to Andrey Loskutov from comment #20) > We've just stumbled upon the table editing regression on GTK3 in 4.7.2 and > I've bissected SWT to this bug which *fixes* the problem for us. > > Eric, I guess this will be too late for a call for backporting this patch to > 4.7.3? Yes, sadly it's too late as we are already into RC4. Besides with these more complex Tree/Table fixes I think it's important to let the fix sit in master for awhile to make sure there are absolutely no regressions.
Verified in I20180305-2000.
Hi Eric, we are trying out the patch here for 4.7.3, and we noticed that some occurrences of the previous fix remain: diff --git a/bundles/org.eclipse.swt/Eclipse SWT/gtk/org/eclipse/swt/widgets/TableItem.java b/bundles/org.eclipse.swt/Eclipse SWT/gtk/org/eclipse/swt/widgets/TableItem.java index 4f3c86a32d..34c91012d8 100644 --- a/bundles/org.eclipse.swt/Eclipse SWT/gtk/org/eclipse/swt/widgets/TableItem.java +++ b/bundles/org.eclipse.swt/Eclipse SWT/gtk/org/eclipse/swt/widgets/TableItem.java @@ -400,9 +400,6 @@ Rectangle getBoundsInPixels (int index) { } int width = OS.gtk_tree_view_column_get_visible (column) ? rect.width + 1 : 0; Rectangle r = new Rectangle (rect.x, rect.y, width, rect.height + 1); - if (parent.getHeaderVisible() && OS.GTK_VERSION > OS.VERSION(3, 9, 0)) { - r.y += parent.getHeaderHeightInPixels(); - } return r; } diff --git a/bundles/org.eclipse.swt/Eclipse SWT/gtk/org/eclipse/swt/widgets/TreeItem.java b/bundles/org.eclipse.swt/Eclipse SWT/gtk/org/eclipse/swt/widgets/TreeItem.java index a3e0e8721f..4cf46f1e88 100644 --- a/bundles/org.eclipse.swt/Eclipse SWT/gtk/org/eclipse/swt/widgets/TreeItem.java +++ b/bundles/org.eclipse.swt/Eclipse SWT/gtk/org/eclipse/swt/widgets/TreeItem.java @@ -453,9 +453,6 @@ Rectangle getBoundsInPixels (int index) { } int width = OS.gtk_tree_view_column_get_visible (column) ? rect.width + 1 : 0; Rectangle r = new Rectangle (rect.x, rect.y, width, rect.height + 1); - if (parent.getHeaderVisible() && OS.GTK_VERSION > OS.VERSION(3, 9, 0)) { - r.y += parent.getHeaderHeightInPixels(); - } return r; } @@ -520,9 +517,6 @@ Rectangle getBoundsInPixels () { } int width = OS.gtk_tree_view_column_get_visible (column) ? rect.width + 1 : 0; Rectangle r = new Rectangle (rect.x, rect.y, width, rect.height + 1); - if (parent.getHeaderVisible() && OS.GTK_VERSION > OS.VERSION(3, 9, 0)) { - r.y += parent.getHeaderHeightInPixels(); - } return r; } Maybe those should be removed. I don't have concrete negative impacts on 4.8, but I definitely see some on 4.7.3. Best regards, Simeon
(In reply to Simeon Andreev from comment #23) > Hi Eric, > > we are trying out the patch here for 4.7.3, and we noticed that some > occurrences of the previous fix remain: > > diff --git a/bundles/org.eclipse.swt/Eclipse > SWT/gtk/org/eclipse/swt/widgets/TableItem.java > b/bundles/org.eclipse.swt/Eclipse > SWT/gtk/org/eclipse/swt/widgets/TableItem.java > index 4f3c86a32d..34c91012d8 100644 > --- a/bundles/org.eclipse.swt/Eclipse > SWT/gtk/org/eclipse/swt/widgets/TableItem.java > +++ b/bundles/org.eclipse.swt/Eclipse > SWT/gtk/org/eclipse/swt/widgets/TableItem.java > @@ -400,9 +400,6 @@ Rectangle getBoundsInPixels (int index) { > } > int width = OS.gtk_tree_view_column_get_visible (column) ? > rect.width + 1 : 0; > Rectangle r = new Rectangle (rect.x, rect.y, width, rect.height + 1); > - if (parent.getHeaderVisible() && OS.GTK_VERSION > OS.VERSION(3, 9, > 0)) { > - r.y += parent.getHeaderHeightInPixels(); > - } > return r; > } > > diff --git a/bundles/org.eclipse.swt/Eclipse > SWT/gtk/org/eclipse/swt/widgets/TreeItem.java > b/bundles/org.eclipse.swt/Eclipse > SWT/gtk/org/eclipse/swt/widgets/TreeItem.java > index a3e0e8721f..4cf46f1e88 100644 > --- a/bundles/org.eclipse.swt/Eclipse > SWT/gtk/org/eclipse/swt/widgets/TreeItem.java > +++ b/bundles/org.eclipse.swt/Eclipse > SWT/gtk/org/eclipse/swt/widgets/TreeItem.java > @@ -453,9 +453,6 @@ Rectangle getBoundsInPixels (int index) { > } > int width = OS.gtk_tree_view_column_get_visible (column) ? > rect.width + 1 : 0; > Rectangle r = new Rectangle (rect.x, rect.y, width, rect.height + 1); > - if (parent.getHeaderVisible() && OS.GTK_VERSION > OS.VERSION(3, 9, > 0)) { > - r.y += parent.getHeaderHeightInPixels(); > - } > return r; > } > > @@ -520,9 +517,6 @@ Rectangle getBoundsInPixels () { > } > int width = OS.gtk_tree_view_column_get_visible (column) ? > rect.width + 1 : 0; > Rectangle r = new Rectangle (rect.x, rect.y, width, rect.height + 1); > - if (parent.getHeaderVisible() && OS.GTK_VERSION > OS.VERSION(3, 9, > 0)) { > - r.y += parent.getHeaderHeightInPixels(); > - } > return r; > } > > Maybe those should be removed. Okay I'll look into it. > I don't have concrete negative impacts on 4.8, but I definitely see some on > 4.7.3. Which ones do you see on 4.7.3 that are not on 4.8?
When we take the patch for our 4.7.3 SWT, SWT cell editors are displaced by the header size. So they are usually shown 1 row below their actual position.
