| Age | Commit message (Collapse) | Author |
|---|
|
|
|
Otherwise, the old status will be retained, which may be non-empty if
it follows an if command whose branch wasn't taken. This is
problematic for scripts using -e.
|
|
Use a buffer bigger than the name limit of devsrv
(127 characters), as this will produce the correct
error message when trying to create the srv file
instead of silently truncating the buffer before.
|
|
The kernel used to always set console=0, which as
usefull during bring-up, but makes it impossible
to use the uart for other purposes.
We now have the ability to pass plan9.ini using
the u-boot script, so add the console=0 line there.
To make debugging easy, we still apply
console=0 if no plan9.ini has been passed.
|
|
mkstemp uses mktemp, which is only declared
when including files with _BSD_SOURCE, so
define _BSD_SOURCE before including the headers
|
|
|
|
|
|
According to the ncurses terminfo database, we have
$ for k in khome kend; do printf '%s:\t' "$k"; TERM=xterm tput $k | od -An -tc; done
khome: 033 O H
kend: 033 O F
$
This differs from vt220fk, so add a new xtermfk table with the proper
sequences.
|
|
Keep track of last hunk offset so we are more likely to find future
hunks.
Start search within the range of lines where we could possibly find
the hunk. This fixes a bug where if the file shrunk since the patch
was generated, and we start the search past the end of the file and
immediately abort the search.
Add an extra offset for the end of the file to the lines array so that
lines[nlines] is the length of the file. This way, when we start the
search at line nlines-oldcnt, we don't access past the end of the array
if oldcnt==0.
When we find a hunk, set lastln to ln + h->oldcnt; we can't apply
another hunk that starts before the previous hunk ends.
|
|
The RFC says that the message separator line consists of
> a timestamp indicating the UTC date and time when the message
> was originally received, conformant with the syntax of the
> traditional UNIX 'ctime' output sans timezone (note that the
> use of UTC precludes the need for a timezone indicator);
It also references http://qmail.org/man/man5/mbox.html as an
authoritative source, which says the date "always contains exactly 24
characters in asctime format".
Add this date format for compatibility with tools using the standard
mbox format, in particular git format-patch.
|
|
The readmessage loop clears errstr at start and expects it not to
change unless there is a read error. However, strtotm may use
multiple calls to tmparse while trying to determine the date format,
which may leave errstr non-empty on success. Clear it after chkunix
so that this doesn't get treated as a message read error.
This also fixes handling of naked From lines, which were previously
just warned about, but since the conversion to tmparse they
accidentally triggered a message read failure.
|
|
it's often not obvious what date component caused an error
when eyeballing a date format string and date; make the
error message contain this information.
|
|
ie NOMK=(gs games) @{ cd /sys/src && mk install }
|
|
|
|
|
|
|
|
|
|
when we have more processors than submit queues,
the submit queues are shared between some processors
and hence we have to acquire its lock.
in the case of the submit queue being full, we have
to unlock it again before retrying otherwise we have
a deadlock.
sorry :(
|
|
Previously, hunks that end exactly at the end of the file could not be
found when the matching lines were located at a positive offset.
|
|
The Readblock enum does not work when you try to read audio cds.
4 cdda blocks are lager than 8192 and the command fails (at least
when using an usb cd drive). This sets the block count according
to the track block size. With this and the fix for libdisk I can
rip audio cds on my el cheapo usb dvd drive. Tested with
"Midnight Oil - Blue Sky Mining"
|
|
0 is a valid mid so should not be used as a nil value, since then the
first mount (mid=0) will never be unmounted.
|
|
Discs and circles were drawn on screen directly and later erased by
offscreen buffer.
This change puts them in line with other operations, which draw to
offscreen first and to screen when necessary
|
|
git/walk had some confusing behavior, where git/add
of a file and then immediate removal would leave
it in the list of files as 'removed'.
This was harmless, but confusing to users. Now,
'git/add' and 'git/rm' without a commit in between
will not list the file as removed.
|
|
Some ethernet drivers like ethervt6105M maintain
a custom pool of Blocks using Block.free
callback.
To maintain that Block.list pointer is
nil when reused, we have to clear it before
returning it to the custom pool as the custom
pool code is not aware of it.
Also, poison Block.list pointer before free().
|
|
Triggered these from ktrans a couple times when poking at it, the
issue is that we get stuck wanting to emit to the tap without a valid
listener. This ensures we are never stuch sending to a deleted window
or dead tap user.
|
|
|
|
|
|
duration
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
As it is implemented right now, OGG pages are not treated properly,
so in case of huge embedded pictures inside metadata the file could
end up ignore by audio/readtags and audio/mkplist. Workaround by
ignoring tags that span across pages for now.
|
|
While the logic described in these has been
preserved carefully and still accurate,
the man page now covers all that is presented
and includes documentation on newer features.
|
|
Revisiting the man page example and README from the
original ktrans there was some descripency on how
to handle punction marks and special characters.
Notably 。(.) and 、(,) need special casing. If we
have anything in the buffer we need to tack on
these characters as an Okuri tail and avoid the actual
punction from entering the Kanji buffer.
Newlines were previously modified to be taken as a completion
if there was runes in the buffer. This has been backed out,
instead Shift + Space can serve this role via kbmap should they
prefer (and as is done in the jp kbmap). Instead we treat newlines
as hints to reset the buffer.
There was also a bug in where after cycling through all options the
original hiragana was not printed back. This has been corrected.
The max number of candidates has been bumped to 32 and moved to an enum.
This does nearly double our resident memory size, but we reguarly had
matches exceed this limit. A better solution is slated.
The man page now makes an attempt to explain the rules around Okuri and
Joshi input modes.
|
|
Apparently 60ms isn't always enough.
|
|
This restores the window image when exiting a subrio.
|
|
We have to ensure the size we compute for memmove(...) in `Expunge` is
properly bounded. For certain combinations of inputs we compute an
illegal size causing a suicide:
upas/fs: imap: fetchrsp: fetchrsp: bad idx 7
upas/fs: user: igor; note: sys: trap: fault read addr=0x0 pc=0x214f92
fs 531: suicide: sys: trap: fault read addr=0x0 pc=0x214f92
Stack trace including state of data-structures:
term% acid -l /sys/src/cmd/upas/fs/imap.acid 531
/proc/531/text:amd64 plan 9 executable
/sys/lib/acid/port
/sys/lib/acid/amd64
acid: lstk()
memmove(p2=0x8e5928,n=0xffffffffffffffe8)+0x42 /sys/src/libc/amd64/memmove.s:36
imap4resp0(imap=0x423c80,mb=0x0,m=0x0)+0x448 /sys/src/cmd/upas/fs/imap.c:522
unexp=0x58b00000000
p=0x425d17
ep=0x425d17
line=0x425d0b
n=0x425d100000058c
verb=0x425d10
op=0x2e6d6f63
imap4resp()+0x1b /sys/src/cmd/upas/fs/imap.c:556
imap4read(imap=0x423c80,mb=0x4239e0)+0x30 /sys/src/cmd/upas/fs/imap.c:928
s=0x425d29
f=0x425d29
n=0x58b0000058c
ll=0x5008a34b0
i=0x425d290000058b
m=0x171bd1f85150e2a6
imap4sync(mb=0x4239e0)+0x62 /sys/src/cmd/upas/fs/imap.c:1059
imap=0x423c80
err=0x0
syncmbox(mb=0x4239e0,doplumb=0x70616d6900000001)+0x5c /sys/src/cmd/upas/fs/mbox.c:76
a=0x58f
n=0x400bc800000000
d=0x0
y=0x0
next=0x171bd206499a9366
m=0x66939a4906d21b17
reader()+0xde /sys/src/cmd/upas/fs/fs.c:1488
t=0x204ca163404153
mb=0x692e6d68656f622f
io()+0x212 /sys/src/cmd/upas/fs/fs.c:1416
n=0x0
main(argc=0x0,argv=0x7fffffffef58)+0x31e /sys/src/cmd/upas/fs/fs.c:353
mboxfile=0x7fffffffef7a
nodflt=0x300000000
srvpost=0x0
v=0x7fffffffef38
_argc=0x6d
_args=0x40d7bd
p=0x400000003
maildir=0x0
mbox=0x0
srvfile=0x0
_main+0x40 /sys/src/libc/amd64/main9.s:15
acid: Imap(0x423c80)
mbox 0x0000000000428b90
freep 0x0000000000423c20
host 0x0000000000423c27
user 0x0000000000423c36
refreshtime 60
cap 0x00
flags 0x05
tag 872
validity 605040277
newvalidity 605040277
nmsg 1419
size 0
f 0x00000000008dd408
nuid 1420
muid 1420
Biobuf bin {
Biobufhdr {
icount -28
ocount 0
rdline 16
runesize 0
state 1
fid 10
flag 0
offset 280380
bsize 8192
bbuf 0x0000000000423d35
ebuf 0x0000000000425d35
gbuf 0x0000000000425cd1
errorf 0x0000000000000000
iof 0x0000000000228d17
aux 0x0000000000000000
}
b end+0x388
}
Biobuf bout {
Biobufhdr {
icount 0
ocount -8192
rdline 0
runesize 0
state 2
fid 10
flag 0
offset 33362
bsize 8192
bbuf 0x0000000000425d9d
ebuf 0x0000000000427d9d
gbuf 0x0000000000427d9d
errorf 0x0000000000000000
iof 0x0000000000228d3a
aux 0x0000000000000000
}
b 0x425d98
}
binit 1
fd 10
The root cause is an integer underflow where we subtract -1 from 0
using an unsigned type.
The above acid trace shows the following values for key local variables:
• nmsg ... 1419
• muid ... 1420
• n ... 1420
• idx ... 1419
The key section of code is /sys/src/cmd/upas/fs/imap.c:516,522
case Expunge:
if(n < 1 || n > imap->muid){
snprint(error, sizeof(error), "bad expunge %d (nmsg %d)", n, imap->nuid);
return error;
}
idx = n - 1;
memmove(&imap->f[idx], &imap->f[idx + 1], (imap->nmsg - idx - 1)*sizeof(imap->f[0]));
Plugging the above values into the call to memmove(...) demonstrates the
issue:
memmove(&imap->f[1419], &imap->f[1419 + 1], (1419 - 1419 - 1)*sizeof(imap->f[0]));
^^^^^^^^^^^^^^^
The third argument of memmove(...) is an unsigned size type causing
the size to be a value that is way too large (the stack trace shows
the size to be the unsigned value 0xffffffffffffffe8).
The `Expunge` case can be fixed by amending the bounds checking
condition before the memmove(...):
case Expunge:
if(n < 1 || n > imap->muid || (n - 1) >= imap->nmsg){
^^^^^^^^^^^^^^^^^^^^^
snprint(error, sizeof(error), "bad expunge %d (nmsg %d)", n, imap->nuid);
return error;
}
idx = n - 1;
memmove(&imap->f[idx], &imap->f[idx + 1], (imap->nmsg - idx - 1)*sizeof(imap->f[0]));
To add some additional context, this issue has been introduced in
revision:
term% git/export 84c4c81ceecfa8f51949787fc2dbe7b14164a353
Date: Mon, 02 Dec 2019 01:12:19 +0000
Subject: [PATCH] upas/fs imap fixes and improvements
do incremental imap fetches after startup, fixes validity handling,
record flags correctly when we aren't in the process of directly
updating a message, fixes off by one in flag parsing, fixes
mis-indexing messages in sync when we get an unsolicited fetch
response.
...
|
|
Graphical display shows current candidate list.
|
|
|
|
|
|
ourselves
|
|
|
|
|
|
|
|
$gitroot has a trailing slash only when the repository is at /. When
it is not, this path construction doesn't work. Use $gitfs/... like
the other git scripts.
|
