Activity for Mamoru TASAKA
-
Mamoru TASAKA posted a comment on ticket #314
For me the problem seems fixed with 2.1.5, thank you.
-
Now I tried Fedora 43 (LXDE) and yes, this crash is reproducible on Fedora 43.
-
I tried 1. i.e. but if you try adding another one Xfe crashes. and crash is reproducible, on Fedora 44: /usr/include/c++/16/bits/stl_vector.h:1253: constexpr std::vector<_Tp, _Alloc>::reference std::vector<_Tp, _Alloc>::operator[](size_type) [with _Tp = FX::FXMenuCommand*; _Alloc = std::allocator<FX::FXMenuCommand*>; reference = FX::FXMenuCommand*&; size_type = long unsigned int]: Assertion '__n < this->size()' failed. Program received signal SIGABRT, Aborted. __pthread_kill_implementation (threadid=<optimized...
-
on xfi window, click right arrow button (click "Previous" button) Oh, this is on xfi window, click left arrow button (click "Previous" button), then perhaps popup window with "Unsupported type: txt" appears, sorry...
-
Perhaps recompiling xfe and fox with sanitizer , e.g. export CC="gcc -fsanitize=address -fsanitize=undefined" export CXX="g++ -fsanitize=address -fsanitize=undefined -fno-sanitize=vptr" or using valgrind makes this issue easier to reproduce.
-
Tentative workaround: diff --git a/src/XFileImage.cpp b/src/XFileImage.cpp index f5beb1d..14faf4b 100644 --- a/src/XFileImage.cpp +++ b/src/XFileImage.cpp @@ -1618,6 +1618,10 @@ FXbool XFileImage::loadimage(const FXString& file) FILE* fp = fopen(file.text(), "r"); + // Save old image + FXImage *img_old = img; + FXImage *tmpimg_old = tmpimg; + if (!fp) { MessageBox::error(this, BOX_OK, _("Error"), _("Unable to open file: %s"), file.text()); @@ -1628,18 +1632,6 @@ FXbool XFileImage::loadimage(const...
-
Use-after-free when xfi goes to invalid image file then trying to redraw it
-
This is already in git: https://github.com/lxde/lxsession/commit/0eb1a31843691b4e5f3354c153315dcb7ebeda3d
-
Note that I have enabled asan (i.e. compiled with gcc -fsanitize=address ). "Normal" compilation (i.e. without asan) may not releave this segfault.
-
make check passes null pointer passed to memcpy with ngspice 43
-
Fix one byte ahead access in dup_string
-
misc/string.c: fix one byte ahead access in dup_string
-
segfault on check with ngspice-43
-
tclspice-42 compilation error with gcc14
-
There is Fedora side downstrem bug: https://bugzilla.redhat.com/show_bug.cgi?id=2252679 and actually the provided commit fixes this issue.
-
Fedora 39 glibc is 2.38 .
-
Apparently specifying more size than the size of destination buffer is not right. Suggested patch: --- xfe-1.45/src/FontDialog.cpp.debug 2022-12-23 20:05:51.000000000 +0900 +++ xfe-1.45/src/FontDialog.cpp 2023-11-14 22:47:01.250243819 +0900 @@ -152,7 +152,7 @@ preview->setBackColor(getApp()->getBackColor()); FXString fontname = FXString(DEFAULT_NORMAL_FONT); - strlcpy(selected.face, fontname.before(',').text(), sizeof(selected.face) + 1); + strlcpy(selected.face, fontname.before(',').text(), sizeof(selected.face));...
-
Okay, now I've received the same bug report on Fedora: https://bugzilla.redhat.com/show_bug.cgi?id=2208727 and actually, when make it sure that pkexec is in the path launch xfe from xterm (or lxterminal) on xfe gui, make it sure that "Authentication using pkexec" is selected on xfe gui try "tool -> new root window", actually I see: "Refusing to render service to dead parents.", while when launching xfe from "gnome-terminal", new root window with pkexec succeeds. Then the problem here is that If yes,...
-
Okay, looks like 1.44 was working. Comparing with 1.44 source code, the following seems to work. --- xfe-1.45/src/WriteWindow.cpp.window_delete 2022-12-23 20:05:50.000000000 +0900 +++ xfe-1.45/src/WriteWindow.cpp 2023-05-06 16:30:24.903053616 +0900 @@ -1532,8 +1532,6 @@ long WriteWindow::onCmdNew(FXObject*, FX window->create(); window->raise(); window->setFocus(); - - delete window; return(1); } @@ -1623,8 +1621,6 @@ long WriteWindow::onCmdOpenRecent(FXObje } window->raise(); window->setFocus();...
-
xfw window disappears after choosing recent files
-
Proposal patch: https://github.com/lxde/lxappearance-obconf/pull/4
-
backtrace: [tasaka1@localhost ~]$ gdb --args lxappearance GNU gdb (GDB) Fedora Linux 12.1-6.fc37 Copyright (C) 2022 Free Software Foundation, Inc. License GPLv3+: GNU GPL version 3 or later <http://gnu.org/licenses/gpl.html> This is free software: you are free to change and redistribute it. There is NO WARRANTY, to the extent permitted by law. Type "show copying" and "show warranty" for details. This GDB was configured as "x86_64-redhat-linux-gnu". Type "show configuration" for configuration details....
-
Reproducible
-
Make compile with -Werror=format-security
-
So looking at the relevant changes on igraph: https://github.com/igraph/igraph/commit/aa7ccccaaeea837d6b9d9b9f2edbf9a759a316f8 https://github.com/igraph/igraph/commit/3f7cdd40cea206aac8c12b782c4a8b76f19395b1 https://github.com/igraph/igraph/commit/7c92169e138cf351da513f8a1e407170a2032c02 I think the following change should do the same behavior as before: --- rw-0.9/simplerw.c.newigraph 2017-02-14 00:20:35.000000000 +0900 +++ rw-0.9/simplerw.c 2022-09-11 19:39:47.033917305 +0900 @@ -134,7 +134,11...
-
Fix for big endian (v1.0.6)
-
This seems again dup of https://sourceforge.net/p/fourpane/bugs/11/ .
-
Okay. This is https://github.com/lxde/lxsession/pull/27 and https://gitlab.gnome.org/GNOME/glib/-/issues/2361 . glib upstream bug report says this is fixed in 2.68.1. Once I can confirm this, I will close my pull request for LXDE side without being merged.
-
Well, no.... looks like the package name is "libglib-2_0-0"? (I guess the version is 2.68.0)
-
Well, not glibc but glib is my concern.
-
Would you post what "glib" version you are using?
-
Actually, for xfe-1.44, make is okay, but make check fails as the original reporter of this says: [tasaka1@localhost xfe-1.44]$ LANGUAGE=en_US.utf8 make check Making check in intl make[1]: Entering directory '/home/tasaka1/rpmbuild/fedora-specific/xfe/rawhide/xfe-1.44/intl' make[1]: Nothing to be done for 'check'. make[1]: Leaving directory '/home/tasaka1/rpmbuild/fedora-specific/xfe/rawhide/xfe-1.44/intl' Making check in po make[1]: Entering directory '/home/tasaka1/rpmbuild/fedora-specific/xfe/rawhide/xfe-1.44/po'...
-
Actually this causes lxappeance segfault when trying to remove icon theme. Forwarded: https://github.com/lxde/lxappearance/pull/4
-
Well, actually if xrandr output is as this, lxrandr should segfault: The problem is preferred 60.00*+ line - current lxrandr source fails to parse this correctly (by regex), however while I want to reproduce this, I don't know how xrandr actually return preferred string....
-
ngspice 34: ngsconvert cannot be built
-
As lxrandr 0.3.2 is released with containing this fix, I think this can be closed.
-
could it be it loads only partially then? I don't think so, actually. Currently, (in most cases) "row-inserted" signal (when loading one desktop file) is emitted after g_signal_connect("row-inserted") finished, so it is going well (although it is in race condition). If some desktop files are loaded before g_signal_connect() is done, the corresponding on_row_inserted() function is not called. So in this case, in load_items(), fm_folder_model_get_item_userdata() should return NULL for those desktop...
-
I want to change "Milestrone"... This should be in "1.3" or "patches" Group.
-
connect_model: connect to signal before setting folder for model
-
Thank you for merging this patch into master.
-
For my usage, now compilation passes with this patch.
-
Fix compilation with gcc10 wrt multiple definitions issue
-
Fix compilation with gcc10 wrt multiple definitions issue
-
Confirmed that this issue is fixed by db91e19e36822432156029d240ed272f56cac010 . Thank you.
-
Note that there are some other issues with gcc10, which defaults to "-fno-common", so there are some "multiple definition of foo" messages, however I cannot submit patches for the issue unless the above linkage error issue is resolved.
-
--enable-xspice --enable-oldapps fals to compile with 32
-
The attached patch seems to cure this issue. With this patch, pcmanfm no longer hangs with Ctrl-C and we don't see 90 seconds hang with systemd.
-
pcmanfm git head hangs with Ctrl-C on Fedora 32
-
Looking at the attached "_usr_bin_lxpanel.1000.crash" report, looking at the assembly, the crash , this is using lxpanel 0.9.3, and the crash is happening at https://git.lxde.org/gitweb/?p=lxde/lxpanel.git;a=blob;f=plugins/task-button.c;h=3e990689c67de7f46d1283f76ea3042a76a0dbdc;hb=9dbda5c15782596983ae61478c1612262633ae41#l1668 , and register says that win is NULL, so this should be the same as https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=855025 and so I think this is fixed by https://git.lxde.org/gitweb/?p=lxde/lxpanel.git;a=patch;f=plugins/task-button.c;h=dfcc9acca9158c8380f9f9add6dafbdfcf96763d...
-
gcc10 now defaults to -fno-common : https://gcc.gnu.org/gcc-10/changes.html So without this patch, lxtask fails to compile with gcc10. Please review and consider to apply this.
-
suppot compilation with gcc10
-
Support gcc10 compilation
-
tclspice does not compile on ngspice 31
-
Because "lxpanel run" just tells "lxpanel" process to launch dialog and then exits immediately, the process showing the dialog is "lxpanel" process, so the PATH the launch dialog is using is the one of "lxpanel" process, not lxpanel process (which already exited). You can workaround this by putting the dialog env PATH=/home/yourname/.local/bin tagtime
-
Please consult Fedora side maintainer (i.e. me). Testing updates will be available tomorrow: https://bodhi.fedoraproject.org/updates/FEDORA-2019-2d0527d781
-
Well, if my patch is applied to the upstream git, perhaps lubuntu maintainer can merge the fix into their repository. I would be happy if someone would review my patch.
-
... And actually, with LANG=zh_TW.utf8 or with LANG=en_US.utf8 , Actually when pressing "Alt+Home" or "Alt+Up", evt->keyval is recognized as "isdigit()", while with ja_JP.utf8 the judgment isdigit() fail (so it works with ja_JP.utf8) And the patch from Michael does work - i.e. with the patch, with LANG=zh_TW.utf8 or with LANG=en_US.utf8, "Alt+Home" or "Alt+Up" come to work. So from me +1 for merging the patch anyway.
-
Well, actually now I can reproduce this issue with LANG=zh_TW.utf8 LC_ALL=zh_TW.utf8 or LANG=en_US.utf8 LC_ALL=en_US.utf8 , however this issue is not reproducible with LANG=ja_JP.utf8 LC_ALL=ja_JP.utf8 . Now I will try debugging (but it may take long time...)
-
Well, for me the previous comment is shown as "Post awaiting moderation", and I have not received it via email yet. So it may be preferable that you would repost your message.
-
Well, I cannot reproduce this issue on Fedora My colleague using Arch linux also is unable to reproduce this issue On the other hand, on Fedora some people are reporting this issue And the above people (i.e. bug reporters on Fedora) say the above patch does NOT work. So this should be related to some environment settings (I guess). Hard to debug because I cannot reproduce...
-
Well, I cannot reproduce this issue My colleague using Arch linux also is unable to reproduce this issue On the other hand, on Fedora some people are reporting this issue And the above people (i.e. bug reporters on Fedora) say the above patch does NOT work. So this should be related to some environment settings (I guess). Hard to debug because I cannot reproduce...
-
Well, unfortunately this does not seem to work... Actually preprocessed source of ngspice-30/src/xspice/icm/analog/climit/ifspec.c shows it doesn't (finally) include src/include/ngspice/cm.h .
-
Okay, for now I will use non-shared compiled one for codemodel files. A happy new year, too! (from Japan)
-
Well, actually what I meant -DSHARED_MODULE or not here is actually not from "command like", but it comes from including ./src/include/ngspice/config.h - including config.h determines whether SHARED_MODULE is defined or so.
-
So at least the difference occurs because when compiled with -DSHARED_MODULE, ngspice-29/src/include/ngspice/ngspice.h includes <errno.h> (while with -USHARED_MODULE it doesn't), and including <errno.h> imports some external symbols. </errno.h></errno.h> So for this case, this can be ignored? However as many other files used for creating codemodel files actuall differ, I am not sure if I can ignore all these differences I currently see...
-
And preprocessed file for ngspice-29/src/xspice/icm/analog/climit/ifspec.c with -USHARED_MODULE
-
And preprocessed file for ngspice-29/src/xspice/icm/analog/climit/ifspec.c with -USHARED_MODULE
-
preprocessed file for ngspice-29/src/xspice/icm/analog/climit/ifspec.c with -DSHARED_MODULE
-
Well, (while not tried currently master branch yet), using ngspice-29, the size of resulting codemodel files actually differs between -with-ngshared and -without-ngshared (when I tried compiling) - I will try investigating further...
-
Make codemodel files parallel installable with/without ngshared
-
Today kernel 4.17.2 hit Fedora and now I've tried it on my "real" machine and it is working as before. Maybe you are using LXDE (or linux) on virtualbox or so?
-
Perhaps I don't have right to review patch on codereview site, however patch looks good to me.
-
Suggestion patch attached. Accessing last + 1 element using [] operator is not allowed. Just use .end() member function.
-
3.1 segfaults with --recursive compiled with gcc8 -D_GLIBCXX_ASSERTIONS
-
Ah, okay, thank you for review and fixing my patch!
-
I think the last comment you see (on lxpanel 0.9.3) is the same as https://sourceforge.net/p/lxde/bugs/900/ .
-
I think the attached patch should fix this issue. Would you review this?
-
gdb log
-
lxpanel: monitor may crash when copy/pasting color from somewhere else
-
Maybe the same as https://sourceforge.net/p/lxde/bugs/900/ ?
-
So it seems that menu must be detached explicitly before widget is to be destroyed. The attached patch works for me. Would you review this?
-
gdb result
-
valgrind result
-
Well, the attached mkv movie is reproducible to me. So: launch two different application (lxterminal and gnome-terminal, for example). on lxpanel taskbar, click right button on one of the taskbutton, and choose "close window", then one application finishes. next, click right botton on the left (another) taskbotton Then lxpanel causes segfault.
-
Sorry, the 4th patch was a bit wrong... There was a mistake when refactoring 4th patch. Use this one instead, thank you.
-
4th patch, for main part, to implement desktop (monitor) insertion. check the number of the current monitors check the number of monitors registered pcmanfm internal lists_desktop_per_screen[] variable if the number of monitors actually increased, do initialize and register new FmDesktop element also show desktop icons
-
3rd patch, for bug fix, for the issue that desktop settings for multiple monitors cannot be saved correctly.
-
Second patch for refactoring, to use GList instead of fixed array, so that new FmDesktop pointer can easily be inserted.
-
First patch for refactoring, to split per-monitor initialization into one function, so that that part can be used when monitor is newly inserted.
-
Support new destop insertion
-
Maybe the same issue as https://sourceforge.net/p/lxde/bugs/895/ . If this issue is still reproducible, would you write the result of "$ xrandr" ?
-
Proposal patch2: Handle the case regex failed to match any monitors
-
Proposal patch1: Allow dot on monitor name
-
[lxrandr]segv when monitor name contains dot
-
Proposal patch here
-
Different window iconized with multihead + current monitor only
-
I can confirm this with lxpanel GIT head. The attached patch should fix this issue.
-
Looks like this is fixed in lxpanel 0.9.0 (ref: https://sourceforge.net/p/lxde/bugs/798/ ), while you are using lxpanel 0.8.1. It would be better that you ask the vendor side maintainer to upgrade lxpanel.
-
Okay, thank you for review. I think your idea is much better, and I think the (newly) attached patch should follow your idea. Note that currently with this patch, I left APIs (of symbols) which are already visible outside unchanged, so I added some "wrapper" functions. And this is "thermal" plugin, not "cpu"... sorry.
-
lxpanel/cpu: panel font size configuration not honored
1 >
