maint(linux): free disk space before running autopkgtests
Running the autopkgtests during a Ubuntu packaging build GHA recently started to fail with an out-of-disk-space error. This change removes some large directories with tools that we don't need for the autopkgtests.
We could delete some more that are not quite as big, but this gives us more than enough space and saves a bit of build-time.
Fixes: #15171
Running the autopkgtests during a Ubuntu packaging build GHA recently
started to fail with an out-of-disk-space error. This change removes
some large directories with tools that we don't need for the autopkgtests.
Fixes: #15171
Build-bot: skip
Test-bot: skip
fix(web): improve determination of default path
Browser extensions can dynamically insert scripts into a web page, so it's possible that the last script is not what we expected, which can cause the path to the script to be wrong. This change fixes the determination of the default path by using the `document.currentScript` property. It also replaces the use of the deprecated `substr` function in those places.
Fixes: #15158
chore(linux): use `devel` for next Ubuntu release
This change moves to use [`devel`](https://hub.docker.com/_/ubuntu#whats-in-this-image) instead of a specific release name for the next Ubuntu release. That way it will always be able to find an image and will automatically start using the next one as soon as it becomes available.
Follow-up-of: #15137
Browser extensions can dynamically insert scripts into a web page, so
it's possible that the last script is not what we expected, which
can cause the path to the script to be wrong. This change fixes the
determination of the default path by using the `document.currentScript`
property. It also replaces the use of the deprecated `substr` function
in those places.
Fixes: #15158
Test-bot: skip
This change moves to use
[`devel`[1]](https://hub.docker.com/_/ubuntu#whats-in-this-image)
instead of a specific release name for the next Ubuntu release. That way
it will always be able to find an image and will automatically start
using the next one as soon as it becomes available.
Build-bot: skip
Test-bot: skip
Adds a call to map2pdb for Delphi projects, if map2pdb is an executable
on the path. This way, we get .pdb files we can use for debugging and
for performance profiling. This also replaces tds2dbg.
If map2dbg is available, then the build and install actions will copy
the PDB to be alongside the corresponding executable, making most debug
actions relatively seamless for finding symbols.
Also fixes up setup.exe in Windows and Developer to extend setup.exe to
a 512-byte boundary after map2pdb patches the PE.
Relates-to: #14787
Build-bot: skip release:developer,windows
Test-bot: skip
Following investigation, it seems that OutputDebugString can trigger a
loader lock as it throws an exception which can be handled by a global
exception handler, which may try and load/unload DLLs or perform other
unsafe tasks (e.g. StackWalk). The safest solution is to remove the
debug string logging altogether from code paths that run from DllMain.
It would be nice to further simplify DllMain but that would be a
significant amount of additional work.
This change wraps the stack walk with an exception handler, which will
enable some crash handlers to continue. However, in some circumstances,
such as the stack captured below, it appears to be an unrecoverable
crash.
So, to mitigate this scenario, we also now only capture stack traces for
crashes that we know are serious -- the set of exception types listed in
the patch.
I experienced this issue on my machine after a Windows update; it is not
clear to me if this is related to the Windows update or if it is due to
my debugging environment for Keyman. The symptoms were that Keyman
Configuration would start up, but show just a blank white dialog, and a
few seconds later would abort. Observing process history showed that
kmshell.exe and kmbrowserhost.exe would both crash, with related call
stacks -- see below for a sample. These call stacks are coming out of
keyman32 calling OutputDebugString from DllMain, which may not be a safe
thing to do, because of the exception side-effect. A separate patch will
be submitted to address that - probably just removing any
OutputDebugString calls from DllMain paths.
In any case, this change adds robustness and should mitigate other
unexpected exceptions.
```
0:026> kb
# ChildEBP RetAddr Args to Child
00 08a1d4c8 750d6e3d 750d6e20 750bd307 04447396 ucrtbase!abort+0x31
01 08a1d4d0 750bd307 04447396 051caaa0 629f52f0 ucrtbase!_purecall+0x1d
02 08a1d508 62ab2914 0495e1c8 00000000 08a1d5fc ucrtbase!__crt_state_management::wrapped_invoke<int (__cdecl*)(void),int>+0x2a
03 08a1d528 62aba00e 0495e1c8 00000000 08a1d5fc dbghelp!SymbolDataSimpleImpl<1,3>::getData+0x24
04 08a1d580 62ac649d 0495e1c8 08a1d5fc 08a1da34 dbghelp!GetData::getSymData+0x6e
05 08a1d5dc 62ac3272 08a1d5fc 05497580 08a1da5c dbghelp!CPubByAddrTrav::get+0xad
06 08a1d948 62ac7340 08a1da00 08a1da34 08a1da5c dbghelp!findBetterSymbol+0x62
07 08a1da40 62ac87eb 08a1da5c 08a1da64 62a05ea0 dbghelp!CAllSymsByAddrTrav::getEnclosingSymbol+0x80
08 08a1da6c 62abfb10 00000001 00030ed1 05497580 dbghelp!CAllSymsByAddrTrav::init+0xeb
09 08a1da88 62a876e9 00000001 00030ed1 043e9008 dbghelp!CAllSymsByAddrTrav::FInit+0xa0
0a 08a1daac 62a87846 043e9008 00000001 00030ed1 dbghelp!CDiaSession::findSymbolByAddr+0x2a9
0b 08a1dadc 62a87899 043e9008 00031ed1 00000000 dbghelp!CDiaSession::findSymbolByRVA+0x76
0c 08a1db10 62aed8f0 043e9008 00031ed1 00000000 dbghelp!CDiaSession::findSymbolByRVAEx+0x39
0d 08a1dbac 62b1a980 533b1ed1 00000000 00000000 dbghelp!FindSymbolByVA+0x117
0e 08a1dc00 629ed8da 533b1ed1 00000000 93b94d79 dbghelp!diaGetFpoFromAddr+0x6c
0f 08a1dc4c 629ed647 533b1ed1 00000000 93b94dad dbghelp!LookupFunctionEntryX86+0x23d
10 08a1dc98 629ed4c7 ffffffff 533b1ed1 00000000 dbghelp!SymFunctionTableAccess64AccessRoutines+0x167
11 08a1dcb4 629ec7a1 ffffffff 533b1ed1 00000000 dbghelp!FunctionTableAccessRoutineLocal+0x17
12 08a1dcd8 629eba25 533b1ed1 00000000 08a1dd10 dbghelp!DbhStackServices::GetFunctionEntry+0x31
13 08a1dd58 629eb126 00000001 08a1de00 08a1eda0 dbghelp!DbsStackUnwinder::StaticAdjustForNoReturn+0x4a
14 08a1ddbc 629e8a76 08a1ddf8 08a1dea8 08a1dec8 dbghelp!DbsX86StackUnwinder::UnwindAndUpdateInternalContext+0x136
15 08a1deec 629e88b4 08a1eda0 00000005 05482908 dbghelp!DbsStackUnwinder::DoDbhUnwind+0x177
16 08a1df1c 629e87b8 08a1eda0 00000005 05482908 dbghelp!DbsStackUnwinder::DbhUnwind+0x58
17 08a1e038 629e867e 08a1eda0 08a1f000 00725a60 dbghelp!PickX86Walk+0x109
18 08a1ecf4 62af77c9 08a1eda0 08a1f000 00000000 dbghelp!DoUnwindStackFrameUsingServices+0xbb
19 08a1ed4c 62b3ad4b 0000014c ffffffff fffffffe dbghelp!StackWalk2+0x1c9
1a 08a1eebc 2c7667c4 0000014c ffffffff fffffffe dbghelp!StackWalk64+0x8b
WARNING: Stack unwind information not available. Following frames may be wrong.
1b 08a1f2cc 2c765ad4 00000000 04e57318 08a1f32c kmshell+0x3667c4
1c 08a1f2dc 7726369f 08a1f2f8 08a1f428 08a1f3d8 kmshell+0x365ad4
1d 08a1f32c 7725eb04 00000000 5338cc66 08a1fbac ntdll!RtlpCallVectoredHandlers+0xf8
1e 08a1f3c0 7726b66f 08a1f3d8 08a1f428 08a1f3d8 ntdll!RtlDispatchException+0x67
1f 08a1f3c0 753e5004 08a1f3d8 08a1f428 08a1f3d8 ntdll!KiUserExceptionDispatcher+0xf
20 08a1f8ec 754c7ab0 40010006 00000000 00000002 KERNELBASE!RaiseException+0x64
21 08a1fb54 533ab6a2 08a1fbac 5338cc66 08a1fe38 KERNELBASE!OutputDebugStringA+0x50
22 08a1fcb0 533b1ba3 53673110 5338cc66 08a1fe38 keyman32!_OutputThreadDebugString+0x42 [D:\Projects\keyman\app\windows\src\engine\keyman32\K32_DBG.CPP @ 291]
23 08a1fd18 533b1ed2 00000000 5338cc66 08a1fe38 keyman32!ReleaseKeyboards+0x13 [D:\Projects\keyman\app\windows\src\engine\keyman32\keyman32.cpp @ 841]
24 08a1fd70 533b1fad 00000000 5338cc66 08a1fe38 keyman32!UninitialiseProcess+0x22 [D:\Projects\keyman\app\windows\src\engine\keyman32\keyman32.cpp @ 160]
25 08a1fdcc 53431b12 53380000 00000003 00000000 keyman32!DllMain+0x5d [D:\Projects\keyman\app\windows\src\engine\keyman32\keyman32.cpp @ 143]
26 08a1fe10 53431cf0 53380000 00000003 00000000 keyman32!dllmain_dispatch+0xb2 [D:\a\_work\1\s\src\vctools\crt\vcstartup\src\startup\dll_dllmain.cpp @ 281]
27 08a1fe24 77268ff6 53380000 00000003 00000000 keyman32!_DllMainCRTStartup+0x20 [D:\a\_work\1\s\src\vctools\crt\vcstartup\src\startup\dll_dllmain.cpp @ 334]
28 08a1fe44 772a0df3 5338cc66 53380000 00000003 ntdll!LdrxCallInitRoutine+0x16
29 08a1fe68 7722b11c 00000003 00000000 fb42ae4c ntdll!LdrpCallInitRoutineInternal+0x22
2a 08a1feb0 77229812 00000003 00000000 fb42af9c ntdll!LdrpCallInitRoutine+0xae
2b 08a1ff60 77261670 1138e500 00000000 052cbe98 ntdll!LdrShutdownThread+0x1b2
2c 08a1ff74 757c5d50 00000000 757c5d30 08a1ffdc ntdll!RtlExitUserThread+0x30
2d 08a1ff84 7725d6db 052cbe98 fb42af20 00000000 kernel32!BaseThreadInitThunk+0x20
2e 08a1ffdc 7725d661 ffffffff 772a4680 00000000 ntdll!__RtlUserThreadStart+0x2b
2f 08a1ffec 00000000 1138e500 052cbe98 00000000 ntdll!_RtlUserThreadStart+0x1b
```
maint(core): fix option defaults for Meson >= 1.1
Meson 1.1 changed the name of the options file from `meson_options.txt` to `meson.options`. Currently we're stuck with Meson 1.0 as minimum Meson version because we still support Ubuntu 22.04 Jammy, but on some machines a newer version of meson gets installed which causes the options to be ignored.
The first attempt to simply symlink or copy `meson_options.txt` to `meson.options` failed because git on Windows doesn't support symlinks by default, and Meson complains if it finds two options files (#15127).
Fortunately not having an options file is only a problem for Core where we have one option in the options file that defaults to true. The solution implemented in this change is to always pass the option to Meson.
Meson 1.1 changed the name of the options file from `meson_options.txt`
to `meson.options`. Currently we're stuck with Meson 1.0 as minimum
Meson version because we still support Ubuntu 22.04 Jammy, but on some
machines a newer version of meson gets installed which causes the
options to be ignored.
The first attempt to simply symlink or copy `meson_options.txt` to
`meson.options` failed because git on Windows doesn't support symlinks
by default, and Meson complains if it finds two options files (#15127).
Fortunately not having an options file is only a problem for Core where
we have one option in the options file that defaults to true. The solution
implemented in this change is to always pass the option to Meson.
Test-bot: skip
While working on the connection between web's TextStore (OutputTarget) and core's context I tried to understand what the `Deadkey` and `DeadkeyTracker` classes are doing. This lead me to add some unit tests and that showed that there were a few bugs in those classes. This PR fixes these:
- initialize `Deadkey.matched` in c'tor
- fix `Deadkey.equal` method
- verify deadkey exists before removing it in `DeadkeyTracker.remove`
- add unit tests for DeadkeyTracker
