| Commit message (Collapse) | Author | Files | Lines |
|---|
|
This reverts commit 878b3cd27fe83f2b0ff476b884c34d165be0072c.
It's no longer required, since the last commit uses a new wcwidth
implementation which covers the cases added by commit 878b3cd2 .
|
|
This commit adds a new wcwidth implementation at libbb/wcwidth_alt.c,
and uses it instead of the existing implementation when compiling for
windows and CONFIG_LAST_SUPPORTED_WCHAR >= 0x30000 - which is the case
with the unicode configs/mingw64u_defconfig.
The windows-target condition keeps non-windows build unmodified, and
the last supported wchar threshold is a semi-hack to allow switching
between implementations without adding a new config option (the old
code supports codepoints up to 0x2ffff).
The new file wcwidth_alt.c was generated by a new scripts/mkwcwidth,
which prints a wcwidth implementation using latest unicode data from
a local clone of https://github.com/jquast/wcwidth . This repo is the
main python wcwidth implementation, and is maintained and up to date.
Functional differences from the existing implementation:
- Unicode 15.1.0 (latest) with the new version (about 450 ranges of
wide and zero-width codepoints), compared to roughly Unicode 5.0
of the existing code (nearly 20 years old spec, about 150 ranges).
The new spec includes, among others, various wide icons and emojis,
which can now be edited correctly at the shell prompt, have correct
alignment in 'ls', etc.
- The old implementation returns -1 (non-printable) for surrogates,
while the new code returns 1, though this is inconsequential, and
POSIX doesn't care. Also libc implementations vary in this regard.
Technical differences:
- The old version compiles less code/data when the last supported
wchar is smaller, while the new version doesn't. This doesn't
matter because the new version is enabled only for the full range.
- The new version is smaller and relatively straight forward, and
fully automated (generated), so updates to newer spec is trivial.
The old version mixes data, ad-hoc code (tailored to the data),
and preprocessor checks, and is hard to automate updates.
The old version has various forms of 32 and 16 bit data ranges, in
several arrays, while the new version uses single data array with
unified form of 32 bits per range, with two rules:
- A data range can't span Unicode planes (enforced, but unlikely
required, and if yes, code to split ranges would be simple).
- A range can't hold more than 32768 codepoints, so bigger ranges
are split automatically (currently there are 2 such ranges).
Performance wise, the new version should be faster, even with three
times the data ranges. Both versions do effectively at most one binary
search in one Unicode plane data, but the new version finds both
zero-width and wide-width results in this one search, while the old
version only finds zero-width, and to detect wide-width it does an
additional linear series of manual range tests, but since most results
are width 1, this sequence is performed in most (non-ASCII) calls.
In a cursory comparison of the new wcwidth with glibc and musl-libc
(both use O(1) lookup tables), with few bodies of text, we're in the
same ballpark, with typical speed of 60% or better.
Bloat-wise, the new version is about 180 bytes code and 1800 bytes
data. If it had similar number of data ranges as the old code (150),
the new version would be about 200 bytes smaller, but because the
new version has 450 data ranges, it's about 1K bigger.
|
|
Commit 26ba73098 (win32: search PATH for missing Unix-style
executables) extended the cases in which PATH would be searched
for executables specified with a Unix-style path.
A side effect of this change was to pass the Unix-style path as
argv[0] to NOFORK applets. The 'uname' applet examines argv[0]
to determine how to behave. When invoked as '/bin/uname' it
returned unexpected results. Other applets may be similarly
affected.
When a NOFORK applet is invoked in evalcommand(), strip any path.
Costs 16 bytes.
(GitHub issue #392)
|
|
POSIX doesn't allow the '#' comment marker to be escaped, though
some implementations do.
As a non-POSIX extension allow '#' to be escaped with a preceding
backslash.
It isn't necessary to escape '#' in macro expansions or command
lines: these cases are covered by an existing extension. Commit
0aceca867 (make: comments in macro expansions and command lines)
Adds 16-32 bytes.
(pdpmake GitHub issue 38)
|
|
The previous commit unnecessarily duplicated the path returned
from resolve_symlinks(), resulting in a memory leak. Thanks to
avih for spotting this.
Even if we can't canonicalise the path in resolve_symlinks() we
can at least return the result of resolving the trailing symlink.
Moreover, this can be done both when the necessary API is missing
and when the filesystem doesn't support it.
Not perfect, but better than nothing, and there's no cost.
This change gets rid of a handful of realpath-related test failures
on ReactOS and Windows XP.
Some background:
Commit 8ebe81483 returned the raw path when the required API was
missing. This was reverted in commit f902184fa because it caused
an infinite loop (GitHub issue #204). Commit 31467ddfc fixed the
cause of the loop, which is why it's now safe to reintroduce a
fallback return.
(GitHub commit #389)
|
|
Some filesystems don't support the feature required to resolve
symlinks for realpath(3).
In such cases just carry on without resolving symlinks. This is
a trade-off between correctness and convenience.
Adds 16 bytes.
(GitHub issue #389)
|
|
Ensure memory is freed even on early exit.
Saves 0-16 bytes.
|
|
Certain tools which allow disk image files and RAM disks to be
mounted were found to lack a feature used to implement realpath(3).
This resulted in a segfault in the 'su' applet when it was run in
such a virtual filesystem.
'su' uses realpath(3) to canonicalise the current directory. This
is only really required to handle network shares mapped to a drive
letter. (GitHub issue #148)
- If the call to realpath(3) fails for some reason fall back to
using the current directory determined by calling getcwd(3).
- If getcwd(3) fails simply don't pass any directory to the shell
being started by 'su'.
Also, ensure all allocated memory is freed, if required.
(GitHub issue #389)
Adds 16-32 bytes.
|
|
Commits 8e6991733 and b2901ce8e fixed problems with the 'read'
shell built-in when the '-t' option was used. However, they
result in a performance penalty. This pipeline:
seq -w 0 999999 | while read line; do :; done
takes 10 times longer than prior to the changes.
If no timeout is specified don't call poll(2).
Costs 16 bytes in a 32-bit build; 0 in 64-bit.
|
|
There's an upstream bug report:
https://bugs.busybox.net/show_bug.cgi?id=15679
that if BusyBox is built for x86_64 with gcc and the use of OpenSSL
for wget is disabled, failures ensue. A similar issue also affects
the busybox-w32 build with the comparable configuration.
The problem appears to be in the assembly code for the function
sp_256_sub_8_p256_mod(), as forcing the use of the provided C code
seems to fix both upstream and busybox-w32. Since my knowledge of
x64_64 assembler is non-existent, someone else will need to determine
the actual issue. For the moment, just avoid the faulty code.
Saves 32 bytes in the 64-bit build.
(GitHub issue #263)
|
|
A 32-bit build of BusyBox using clang segfaulted in the test
"awk assign while assign". Specifically, on line 7 of the test
input where the adjustment of the L.v pointer when the Fields
array was reallocated
L.v += Fields - old_Fields_ptr;
was out by 4 bytes.
Rearrange to code so both gcc and clang generate code that works.
Signed-off-by: Ron Yorston <rmy@pobox.com>
Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>
|
|
log5() with crondlog(5, msg, va) seems making logging more consistent.
function old new delta
------------------------------------------------------------------------------
(add/remove: 0/0 grow/shrink: 0/0 up/down: 0/0) Total: 0 bytes
Signed-off-by: Jones Syue <jonessyue@qnap.com>
Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>
|
|
function old new delta
subevalvar 1576 1588 +12
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
|
|
I developed this application to test the Linux kernel series [1]. As
described in it I could not use the iproute2 package since the
microcontroller is without MMU.
function old new delta
do_set_can - 920 +920
packed_usage 34645 34908 +263
get_float_1000 - 164 +164
.rodata 105427 105539 +112
do_iplink 1313 1381 +68
------------------------------------------------------------------------------
(add/remove: 2/0 grow/shrink: 3/0 up/down: 1527/0) Total: 1527 bytes
cc: Marc Kleine-Budde <mkl@pengutronix.de>
[1] https://marc.info/?l=linux-netdev&m=167999323611710&w=2
Signed-off-by: Dario Binacchi <dario.binacchi@amarulasolutions.com>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
|
|
function old new delta
parse_stream 2271 2292 +21
.rodata 105408 105427 +19
------------------------------------------------------------------------------
(add/remove: 0/0 grow/shrink: 2/0 up/down: 40/0) Total: 40 bytes
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
|
|
function old new delta
set_G_ifs - 151 +151
run_list 1024 1031 +7
run_pipe 1567 1445 -122
------------------------------------------------------------------------------
(add/remove: 1/0 grow/shrink: 1/1 up/down: 158/-122) Total: 36 bytes
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
|
|
|
|
|
|
function old new delta
.rodata 105412 105408 -4
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
|
|
|
|
Older versions of mingw don't define PROCESSOR_ARCHITECTURE_ARM64.
Don't let this stop the build.
|
|
Commit 992387539 (build system: more clang/llvm tweaks) added a
test in scripts/Makefile.build which used the intcmp function.
This isn't present in GNU make prior to 4.4.
Rewrite the test so it works with older versions of GNU make.
|
|
The last two commits allow EXTRAVERSION to track the current state
of a git repository. The build system was unable to determine
which files were affected by changes to EXTRAVERSION and caused
a full rebuild when it changed.
Alter how the version information is passed to the code so only
a handful of files need to be rebuilt when it changes.
|
|
The previous commit to detect EXTRAVERSION more accurately was
lacking in some respects:
- Only attempt to detect EXTRAVERSION if it's currently set to the
default value '.git'.
- Use immediate-expansion macro assignments to avoid time-consuming
repeated evaluations.
- Remove a useless use of cat.
- Correctly allow for the output being sent to a separate directory.
|
|
The default EXTRAVERSION is .git. Use some clues to determine a
more precise EXTRAVERSION:
- git describe
- A .frp_describe file embedded in the release tarball
- The name of the build directory: tarballs of release tags from
GitLab and GitHub include the version in the directory name.
|
|
The web server in BusyBox has an optional feature to allow pages
with a given suffix to be processed by a script interpreter: PHP,
for example.
Enable this in the busybox-w32 build.
Costs 256-320 bytes.
(GitHub issue #266)
|
|
When setting up the arguments for the CGI handler:
- Use '-I0' rather than two separate arguments.
- Use memcpy() to copy the server arguments.
Saves 32 bytes.
|
|
The httpd server made unnecessary calls to detach from the console.
Avoiding this saves one process creation when a file is being
served and two for a CGI script.
Then again, commit ad4bd5bbd (win32: avoid console windows from CGI
scripts) results in a console host always being created for a CGI
script even if it turns out to be unnecessary. So the net saving
is only one process for CGI scripts.
|
|
Commit ea8742bc16 (ash: workaround environment issue in Windows
on ARM) prevented nofork applets from being run without a fork
in standalone shell mode in Windows on ARM. This was due to a
deficiency in the handling of the 'environ' global.
However, the problem is specific to MSVCRT. If the target is
UCRT nofork applets can be made to work without a fork.
|
|
When httpd is run in the background its processes are detached
from the console. CGI scripts could create subprocesses which
needed a console, resulting in annoying console windows appearing.
Prevent this by changing the creation flags for CGI scripts to
CREATE_NO_WINDOW.
|
|
Update the mkrelease script to include cross-compilation of the
Windows on ARM binary on Linux using llvm-mingw.
|
|
In the Microsoft Windows build stat(1) requires find_mount_point()
from libbb. This won't be available unless df(1) is also enabled.
Add a dependency so stat(1) can be built without df(1). This may
result in find_mount_point() being compiled needlessly in some
upstream builds, but we can live with that.
(GitHub issue #385)
|
|
- Rename some functions to be more meaningful.
- Adjust conditional compilation to clarify which code is required
for 'standalone shell' and 'exec prefers applets' settings.
This shouldn't result in any change to the behaviour or size of
default builds.
|
|
When an unexpected value is detected in UTF-8, we should print the
placeholder codepoint, and then recover whenever we detect a value
which is valid for starting a new UTF-8 codepoint (including ASCII7).
However, previously, we only tested recovery at the bytes following
the unexpected one, and so if the first unexpected value was also
valid for a new codepoint, then didn't rcover it.
Now we check for recovery from the first unexpected byte, which,
if recoverable, requires both placeholder printout and recovery,
so the recovery "unwinding" is modified a bit to allow placeholder.
Example of of a sequence which now recovers quicker than before:
(where UTF-8 for U+1F600 "😀" is: 0xF0 0x9F 0x98 0x80)
printf "\xF0\xF0\x9F\x98\x80A"
Previously: ?A
Now: ?😀A
|
|
The mingw-w64 project has updated its implementation of dirname(3).
In some circumstances the new version doesn't preserve the type of
the user-supplied top-level directory separator. As a result of
this the dirname-handles-root test case failed.
Import the new implementation and tweak it to preserve the type of
the separator.
This only affects mingw-w64 versions 12 and above. Currently only
the aarch64 build using llvm-mingw is affected.
|
|
If the NO_COLOR environment variable is set and is not empty 'ls'
won't output ANSI colour codes. This is an alternative to the
existing approach of setting 'LS_COLORS=none'.
See https://no-color.org/.
Costs 24-32 bytes.
(GitHub issue #382)
|
|
In external_exists() in appletlib.c it's necessary to take a copy
of the pointer to the allocated variable path1 so it can be freed:
find_executable() will change its value.
|
|
It seems windres in llvm doesn't understand MANIFEST resources.
Use the numeric value 24 instead.
|
|
The conditional compilation to control standalone shell mode was
incorrect when building for POSIX. This hadn't been noticed before
as it had only been tested in the default configuration where
standalone shell mode is disabled.
|
|
Use -fsjls-exceptions with clang.
|
|
Commit 992387539 (build system: more clang/llvm tweaks) falsely
claimed that the lack of the --warn-common linker option was being
detected.
It wasn't, but it is now.
|
|
For Windows on ARM we need to report the aarch64 processor
architecture.
|
|
The top-level makefile attempted to detect if make was being run
in silent mode (-s option). It then turned off reporting of
commands as they were run. There were false positives where
unrelated 's' characters in the command line were detected as
silent mode.
Import the relevant section from the latest Linux build system to
improve matters.
|
|
Update the mkprerelease script to include cross-compilation of
the Windows on ARM binary on Linux using llvm-mingw.
|
|
The top-level Makefile hardcoded the host and cross compiler name
to "gcc". Make it possible to set different values either on the
command line (make CROSS_COMPILER=clang) or in the configuration
file.
|
|
A 32-bit build of BusyBox using clang segfaulted in the test
"awk assign while assign". Specifically, on line 7 of the test
input where the adjustment of the L.v pointer when the Fields
array was reallocated:
L.v += Fields - old_Fields_ptr;
was out by 4 bytes.
Rearrange to code so both gcc and clang generate code that works.
This patch has been submitted upstream. Until it's accepted there
the new code is only used in builds for Windows.
|
|
The environment is handled differently in ARM64 Windows.
Assignments to 'environ' result in a compiler error. Omit the
offending code for now. NOFORK applets will cause a fork.
|
