9pfuse fails on ARM when O_LARGEFILE is supported.
glibc does define O_LARGEFILE properly on ARM,
and the value is different than what that this workaround suggests,
causing it to wrongly detect bad flags.
Change-Id: I02b0cc222ca7785c4b1739c3df3caa17cf7bc265
Reviewed-on: https://plan9port-review.googlesource.com/1094
Reviewed-by: Russ Cox <rsc@swtch.com>
Credit to Roi Martin <jroi.martin@gmail.com> for noticing that
libdraw was being passed a negative string length and for finding the
sequence of keystrokes that make acme do it reproducibly.
Change-Id: If3f3d04a25c506175f740d3e887d5d83b5cd1bfe
Reviewed-on: https://plan9port-review.googlesource.com/1092
Reviewed-by: Russ Cox <rsc@swtch.com>
On OSX 10.10, when you open an application that depends on devdraw, the
title bar only shows the first letter of the application's name. The
patch sets a default title as soon as the window is created, which
fixes this issue.
On OSX 10.10, when you open an application that depends on devdraw, this
application is opened in top of other windows, however the menu bar is
not updated. The patch calls topwin() at the end of makewin() in
src/cmd/devdraw/cocoa-screen.m .
Change-Id: Ie036928b5574c8df20ad8b2b54047e2f7a22bb41
Reviewed-on: https://plan9port-review.googlesource.com/1091
Reviewed-by: Russ Cox <rsc@swtch.com>
These were *.C back in 2003, before the 9 script, to avoid conflicts
with the system utilities. A later change renamed them, but that
change seems to have been lost during the hg->git conversion
because I ran the conversion on a case-insensitive file system.
Change-Id: Id32c99cb9571ef0e185c3cc9e8c8d6d5b48ca195
fixed warnings:
src/cmd/fossil/disk.c:37:14: warning: use of GNU 'missing =' extension in designator [-Wgnu-designator]
src/cmd/fossil/disk.c:38:14: warning: use of GNU 'missing =' extension in designator [-Wgnu-designator]
src/cmd/fossil/disk.c:39:14: warning: use of GNU 'missing =' extension in designator [-Wgnu-designator]
src/cmd/fossil/disk.c:40:13: warning: use of GNU 'missing =' extension in designator [-Wgnu-designator]
src/cmd/fossil/disk.c:41:14: warning: use of GNU 'missing =' extension in designator [-Wgnu-designator]
src/libndb/ndbreorder.c:41:55: warning: for loop has empty body [-Wempty-body]
ignored warnings:
src/cmd/acid/dbg.y:393:9: warning: array index -1 is before the beginning of the array [-Warray-bounds]
src/cmd/bc.y:1327:9: warning: array index -1 is before the beginning of the array [-Warray-bounds]
src/cmd/bc.y:1327:9: warning: array index -1 is before the beginning of the array [-Warray-bounds]
src/cmd/grep/grep.y:420:9: warning: array index -1 is before the beginning of the array [-Warray-bounds]
src/cmd/grep/grep.y:420:9: warning: array index -1 is before the beginning of the array [-Warray-bounds]
src/cmd/hoc/hoc.y:692:9: warning: array index -1 is before the beginning of the array [-Warray-bounds]
src/cmd/hoc/hoc.y:692:9: warning: array index -1 is before the beginning of the array [-Warray-bounds]
src/cmd/lex/parser.y:886:9: warning: array index -1 is before the beginning of the array [-Warray-bounds]
src/cmd/rc/syn.y:303:9: warning: array index -1 is before the beginning of the array [-Warray-bounds]
src/cmd/units.y:1003:9: warning: array index -1 is before the beginning of the array [-Warray-bounds]
src/libregexp/regcomp.c:19:16: warning: variable 'reprog' is not needed and will not be emitted [-Wunneeded-internal-declaration]
LGTM=rsc
R=rsc
https://codereview.appspot.com/158250043
Acme tracks the most recent typing insertion point and
the home and end keys stop there on their way
up to the top or down to the bottom of the file.
That point should be iq1, and it should be adjusted
properly so that it's always between 0 and t->file->b.nc inclusive.
(This is all code from an external contributor, years old at this
point but new since Plan 9.)
Somehow, sometimes iq1 ends up a little beyond b.nc,
and when passed to textbacknl it crashes acme in bufread.
I can't see how that can happen but if it does, avoid the crash.
It's tempting to pull the insertion point code out entirely
but this is a little less invasive and should fix things for now.
TBR=rsc
https://codereview.appspot.com/107730043
We ran for a long time with 10ms kernel resolution,
so 10ms user space resolution here should be fine.
Some systems actually provide 1ms sleeps, which
makes this polling use a bit more cpu than we'd like.
Since the timers are for user-visible things, 10ms should
still be far from noticeable.
Reduces acme's cpu usage on Macs when plumber is missing
(and plumbproc is sleeping waiting for it to appear).
LGTM=aram, r
R=r, aram
https://codereview.appspot.com/99570043
This breaks ^C in win windows, as expected.
People use ^C, win expects and handles ^C,
so I don't think we can just take it away.
I've noticed that it is broken but assumed my ssh
was screwed up.
If you want to make WindowsKey+C,X,V do the
operations, by analogy with command+C,X,V
on Mac, that's fine with me.
««« original CL description
acme: copy/cut/paste with ctl+c,x,v
LGTM=rsc
R=rsc
CC=plan9port.codebot
https://codereview.appspot.com/69070045
»»»
TBR=rsc
CC=burns.ethan, r
https://codereview.appspot.com/96410045
smtp.c:232: warning: comparison with string literal results in unspecified behavior
smtp.c:244: warning: comparison with string literal results in unspecified behavior
marshal.c:1179: warning: variable ‘err’ set but not used
LGTM=rsc
R=rsc
https://codereview.appspot.com/93290043
Reading /mnt/acme/log reports a log of window create,
put, and delete events, as they happen. It blocks until the
next event is available.
Example log output:
8 new /Users/rsc/foo.go
8 put /Users/rsc/foo.go
8 del /Users/rsc/foo.go
This lets acme-aware programs react to file writes, for example
compiling code, running a test, or updating an import block.
TBR=r
R=r
https://codereview.appspot.com/89560044
Bakul Shah has observed corrupted files being written
when acme writes over osxfuse to sshfs to a remote file system.
In one example we examined, acme is writing an 0xf03-byte
file in two system calls, first an 0x806-byte write and then a 0x6fd-byte
write. (0x806 is BUFSIZE/sizeof(Rune); this file has no multibyte UTF-8.)
What actually ends up happening is that an 0x806-byte file is written:
0x000-0x6fd contains what should be 0x806-0xf03
0x6fd-0x7fa contains zeros
0x7fa-0x806 contains what should be 0x7fa-0x806 (correct!)
The theory is that fuse or sshfs or perhaps the remote file server is
mishandling the unaligned writes. acme does not seem to be at fault.
Using bio here will make the writes align to 8K boundaries,
avoiding the bugs in whatever underlying piece is broken.
TBR=r
https://codereview.appspot.com/89550043
- the cursor is on the last line
- the navigation would put the cursor over the tag of the following text
R=rsc
CC=smckean83
https://codereview.appspot.com/15280045
Introduces the Search command for mailboxes.
Arguments passed are treated as one space-
separated string, passed on to mailfs' IMAP
search interface.
R=rsc, david.ducolombier
CC=plan9port.codebot
https://codereview.appspot.com/13238044
Mail services (such as Google Mail) will often have
directories with names that contain spaces. Acme
does not support spaces in window names. So, replace
spaces in mail directory names with the Unicode
character for visible space.
The code is a bit of an over-approximation and
generally non-optimal.
R=rsc, david.ducolombier, 0intro
CC=plan9port.codebot
https://codereview.appspot.com/13010048
UTF-8 searches with the SEARCH command must
be conducted in two steps: the first sends
the SEARCH command with the length of the
UTF-8 encoded string and the second sends
the literal search term. The searches need
to not be quoted.
R=rsc, david.ducolombier, rsc, 0intro
CC=plan9port.codebot
https://codereview.appspot.com/13244043
Everyone seems to assume that TERM != dumb implies
ANSI escape codes are okay. In fact, many people assume
that unconditionally, but it is easier to argue back about
TERM=dumb than TERM=9term.
This applies to acme win too, because they share the code.
Set termprog=9term or termprog=win for clients who
need to know.
R=rsc
CC=r
https://codereview.appspot.com/12532043
gcc compiles `p + length < p' into 'length < 0' since pointer overflow is undefined behavior in C. This breaks the check against a large `length'.
Use `length > pend - p' instead.
There's no need to check `length < 0' since `length' is from length_decode() and should be non-negative.
===
Try the simplified code.
void bar(void);
void foo(unsigned char *p, int length)
{
if (p + length < p)
bar();
}
$ gcc -S -o - t.c -O2
...
foo:
.LFB0:
.cfi_startproc
testl %esi, %esi
js .L4
rep
ret
.L4:
jmp bar
.cfi_endproc
Clearly `p' is not used at all.
R=rsc
CC=plan9port.codebot
https://codereview.appspot.com/7231069
Ubuntu Precise seems to have a buggy X server
that sometimes fails at XCopyArea. Let devdraw
do it itself.
This will slow down remote X a little bit,
but slow and correct is better than fast and broken.
R=rsc
https://codereview.appspot.com/7310069
The code is adapted from Plan 9's import(4); this allows us to speak
that protocol. We don't currently support AAN (in the works) or
TLS/SSL.
Thanks to David for help with the man page, testing, and development.
R=0intro, rsc
CC=plan9port.codebot
http://codereview.appspot.com/6458100
If the mouse was in the tag of the old window,
it was most likely pointing at Del. If bringing up a
new window from below and not moving the mouse
somewhere else, adjust it so that it ends up pointing
at Del in the replacement window's tag too.
This makes it easy to Del a sequence of windows in
a column, from top to bottom.
http://www.youtube.com/watch?v=ET8w6RT6u5M
R=r
http://codereview.appspot.com/6558047
Change is to handle FUSE_FORGET in main loop instead of separate thread for each as 10s of thousands can come in at once.
Fixes issue 104.
R=0intro, rsc
http://codereview.appspot.com/6498081
On some systems, the third argument of connect() and bind()
is expected to be the length of the address family instead
of the length of the sockaddr structure.
R=rsc
http://codereview.appspot.com/6489072
In the current code, the srv file is removed
just after the main thread exits, while the
srv thread is still running, which is not
the expected behavior.
We moved the srv creation just before the
procrfork, in order that the srv file will
not be removed until the srv thread exits.
R=rsc
http://codereview.appspot.com/6397047
This patch reverts the recent patch named "devdraw: fix for OS X 10.8", and fixes redrawing on OSX 10.8 differently, making scrolling under Acme 3 times faster.
R=rsc, mirtchovski
CC=plan9port.codebot
http://codereview.appspot.com/6452087
In MacOS 10.8, the NSBitmapImageRep class appears to cache the specified
image data at the time of construction. As a result updates to the
backing memimage object do not get pushed to the screen in flushimg.
This patch creates the NSBitmapImageRep object over again for each
flushimg which would appear to fix the problem.
R=rsc
CC=plan9port.codebot
http://codereview.appspot.com/6443063
The function p9dialparse() returns the host as a sockaddr_storage
structure instead of a u32int, to be able to handle both IPv4
and IPv6 addresses. Because the sockaddr_storage structure also
handle port numbers and Unix path names, there is no longer
need to set them in the calling functions. However, these values
are still returned for convenience.
The sockaddr_in and sockaddr_un structures have been replaced
by sockaddr_storage to handle Unix, IPv4 and IPv6 sockets.
Names and addresses are resolved using either gethostbyname()
or getaddrinfo() functions.
The getaddrinfo() function is documented in RFC2553 and standardized
since POSIX.1-2001. It supports both IPv4 and IPv6 addresses.
The gethostbyname() function is deprecated since POSIX.1-2008.
However, some libc implementations don't handle getaddrinfo()
properly, thus we preferred to try gethostbyname() first.
I've tried to preserve most of the old code logic to prevent
from surprising or unwanted behavior.
R=rsc
http://codereview.appspot.com/6255068
there is no more blank image when toggling
fullscreen, when unminimizing, or at startup; this
also means that we can live resize, but we only
enable live resizing with Page, because it seems
useless for other apps, and Acme and Sam bug with
it. The tradeoff is that bottom corners are
sometimes automatically rounded. There is a way to
prevent the rounding here:
http://parmanoir.com/Custom_NSThemeFrame
but this would obfuscate the code. Instead, we
make sure that the corners are always rounded.
Closing the window while its content is updated
causes an exception, without this patch. This
seems to happen regularly with stats(1).
This patch avoids a possible deadlock at startup,
which I never experienced. If I recollect right,
there is little chance that this happens on a
multi-core CPU.
Minimizing now activates next app in line, and
Devdraw now stops drawing while minimized.
R=rsc
CC=plan9port.codebot
http://codereview.appspot.com/5499043
Now, if you full-screen an app running on a secondary
screen, it only takes over that monitor. It does not cause
the primary screen to disappear.
Code by David Jeannot.
R=djeannot24
http://codereview.appspot.com/5708046
These functions are equivalent to vtSha1 and vtSha1Check
from the old libventi and are particularly used by Fossil.
R=rsc
http://codereview.appspot.com/5555064
Previously, the menu bar has been set to hidden
instead of autohidden in fullscreen mode, because
it appeared too easily. Occasionally however, it
still appears. This patch makes sure it won't.
With OS X Lion's fullscreen, the Dock is
autohidden, but sometimes remains hidden. This
bug comes with visual menu bar bugs
during space transition and when exiting
fullscreen. Those bugs happen after each
Devdraw's activation, for example. This patch
neutralizes them.
R=rsc
CC=plan9port.codebot
http://codereview.appspot.com/5502091
This implementation is simpler to understand. It
removes conflicts with other modifiers and with
the virtual buttons. Pressing alt during a
sequence now cancels it correctly.
R=rsc
CC=plan9port.codebot
http://codereview.appspot.com/5507049
Fix segfault when the first archived file is DMSYMLINK or DMDEVICE.
Reinitialize the buffer to zero before calling readlink.
R=rsc
CC=plan9port.codebot
http://codereview.appspot.com/5500097
(The new variable "willactivate" is in the input
structure "in", which is illogical. But this
structure will soon be renamed "app". I postpone
the renaming to avoid conflicts with simultaneous
Codereview issues.)
R=rsc
CC=plan9port.codebot
http://codereview.appspot.com/5504102
This problem happens especially in fossil/view when vtread
can be called even if not connected to venti.
The old libventi handle this case properly in the function vtRPC, which is called by vtReadPacket (client.c).
R=rsc
CC=plan9port.codebot
http://codereview.appspot.com/5484044
Home and End previously navigated between
two different window locations: the top and
the bottom of the text. Now they include a
third waypoint: the location where typing last
happened. Thus, in a win window, typing
ls -l
<home>
scrolls to the beginning of the ls -l output.
A second <home> continues to the top of the file.
Makes Send scroll always, along with writes by
external programs to +Errors.
R=r
CC=mccoyst
http://codereview.appspot.com/4830051
The scratchbox-based SDK for maemo cross-compilation (for Nokia N800 and N900) has a non-standard uname command which doesn't support the '-p' option. This patch allows INSTALL to work anyway, by issuing 'uname -m' if 'uname -m -p' fails.
R=rsc
CC=plan9port.codebot
http://codereview.appspot.com/4554041
Get p9p to build cleanly for ARM - based on suggestions from rsc and eekee, plus a final tweak from me. Tested (a bit) on sheevaplug (ubuntu jaunty), Nokia N800 (maemo 4) and Nokia N900 (maemo 5).
R=rsc
CC=plan9port.codebot
http://codereview.appspot.com/4524056
Ignore scroll/noscroll window setting.
Instead, scroll when the write begins in
or immediately after the displayed window content.
In the new scrolling discipline, executing
"Noscroll" is replaced by typing Page Up or
using the mouse to scroll higher in the buffer,
and executing "Scroll" is replaced by typing End
or using the mouse to scroll to the bottom of
the buffer.
R=r, r2
http://codereview.appspot.com/4433060
Add -x flag to mklatinkbd to generate file for $HOME/.XCompose.
Thanks to Anthony Martin, Tony Lainson, and Kris Maglione for legwork.
R=rsc
http://codereview.appspot.com/3811043
