Reini Urban wrote:
Thinking some more about this, there are really some inconsistencies with
the current and proposed behavior that I don't like.
[...]
I have no strong opinion in these issues (yet), but please look also at
the related ending-colon ':extension' problem on NTFS.
Such
Brian Dessent wrote:
Let me say it again. This is not new behavior:
2003-08-05 Pavel Tsekov ptsekov AT gmx.net
* path.cc (cygdrive_getmntent): Do not skip over drives of type
DRIVE_REMOVABLE.
Perhaps you should be discussing this with Pavel.
Okay, I
.
Example:
$ /bin/kill -s `perl -e 'print Ax200'`
Segmentation fault (core dumped)
As far as I can tell from CVS history this has existed in kill.cc since
its first version (~5 years.) Trivial patch below.
2005-02-26 Brian Dessent [EMAIL PROTECTED]
* kill.cc (getsig): Use
placeholder
value that is initialized to .. paths[1] contains the CWD anyway so
there's no need to examine paths[0].
Brian
===
2005-03-24 Brian Dessent [EMAIL PROTECTED]
* cygcheck.cc (init_paths): Use full path instead
getservbyname(auth, tcp) success, port = 113
getservbyname(25, tcp) success, port = 113
Brian
===
2005-04-05 Brian Dessent [EMAIL PROTECTED]
* net.cc (__dup_ent): Make dst point to NULL if src is NULL.
Free dst
Christopher Faylor wrote:
Thanks for the patch, but I went out of my way to avoid freeing the
buffer when I maded changes to dup_ent a couple of weeks ago. I don't
want to revert to doing that again, so I've just used the return value
in all cases.
Thanks for taking care of that. My
This is just a trivial change of argument to execl() testcases, which
supresses the warning 'missing sentinel in function call' in gcc4 that
causes the tests to fail.
winsup/testsuite
2005-05-17 Brian Dessent [EMAIL PROTECTED]
* winsup.api/signal-into-win32-api.c (main): Use 'NULL
Christopher Faylor wrote:
Go ahead and check these in but please use GNU formatting conventions,
i.e., it's (char *) NULL, not (char *)NULL. Actually, isn't just NULL
sufficient?
I must have had C++ on the mind, thinking that the cast was necessary.
Sorry but no. This is a workaround. We
of steps that should cover everything.
2005-05-17 Brian Dessent [EMAIL PROTECTED]
* install.texinfo (How do I uninstall...): Rewrite to cover
removing services, dealing with permissions, and other common
tasks for removing Cygwin completely.Index: install.texinfo
Corinna Vinschen wrote:
2005-05-17 Brian Dessent [EMAIL PROTECTED]
http://cygwin.com/acronyms#PCYMTNQREAIYR ;-)
Yeah, I know. Spammers have had my address for some time, I don't feel
like hiding. Me heart SpamAssassin. :)
Close all Cygwin command prompts, xterms, etc. and stop
Corinna Vinschen wrote:
Alright. I am not sure how to push out the new version to the web site,
so someone else will have to do that (or tell me what to do - check in
the .html files into the website CVS or something?)
Yep. cvs -d :ext:cygwin.com:/cvs/cygwin co htdocs
Got it, thanks.
Corinna Vinschen wrote:
While this might help to avoid... something, I'm seriously wondering
what's wrong with this expression. Why does each new version of gcc
add new incompatibilities?
I think I've figured this out. PR/13684 added thread safety to
initialization of local statics.[1] It
this.
Brian
2005-05-19 Brian Dessent [EMAIL PROTECTED]
* cygrunsrv.cc (longopts): Add '--list' and '--verbose' options.
(opts): Add '-L' and '-V' options; keep order consistent with above.
(action_t): Add 'List'.
(err_out_set_error): Define version of 'err_out' macro
Brian Dessent wrote:
-controlsToString(DWORD controls)
+ char *base, *end;
+ static char buf[34];
+ int used = 0, dsiz = strlen (delim);
Crap, that is a mistake. buf[34] should be something more generous like
128 or 256. I had it set small to test to make sure it couldn't
overflow
or not but I would imagine most direct
references to Cygnus are probably anachronisms.
Brian
2005-05-22 Brian Dessent [EMAIL PROTECTED]
* cygcheck.cc (dump_sysinfo_services): Add new function that uses
new cygrunsrv options to dump service info.
(dump_sysinfo): Call
) if there is a problem connecting or resolving the domain.
Brian
winsup/utils:
2005-06-20 Brian Dessent [EMAIL PROTECTED]
* Makefile.in: Link cygcheck with libwininet.a.
* cygcheck.cc: Add includes.
(grep_packages): New global variable.
(display_internet_error): New
Christopher Faylor wrote:
Btw, the other license provision in the cygwin licensing web page was
really meant as a way to accommodate other, already existing projects.
And it was very gracious of them to do that. For an example of why this
makes life a lot easier, consider MySQL (GPL) and
Brian Dessent wrote:
why not simply run cygrunsrv --list --verbose in verbose mode, instead
of actually going through one iteration of the loop? Simply to reuse the
copy output code? Brian?
The reason I did it that way is because if I had run cygrunsrv --list
Now that I re-read what
Christopher Faylor wrote:
Go ahead and check this in. Thanks.
Ok.
Thanks, Igor, for bringing this up (again).
Thanks Igor, I had meant to bring this up but forgot.
There's no need to ping anybody. I do read this list and I haven't
forgotten about the patch. If it didn't require changs
This uses LoadLibrary and GetProcAddress instead of -lwininet so that systems
lacking IE3 can still run cygcheck. Tested on XP and NT4, and verified that
with WININET.DLL renamed cygcheck can still function.
2006-01-13 Brian Dessent [EMAIL PROTECTED]
* Makefile.in (cygcheck.exe): Do
Christian Franke wrote:
At least when regtool is used interactively, it is IMO not very useful
to have
modem-line-noise-like output for REG_BINARY, but human readable output for
the other value types.
But this is the current behavior of regtool get
Instead of an explicit -b flag,
Igor Peshansky wrote:
What if you want to redirect the hex dump to a file? IMO, isatty() checks
are only useful if the output won't change qualitatively on redirection
(e.g., for coloring). Otherwise, it's always better to use an explicit
flag.
Good point. Why don't we just emulate the
/cygwin:
2006-02-02 Brian Dessent [EMAIL PROTECTED]
* cygtls.h: Include sys/cygwin.h.
(myfault::~myfault): If a debugger is present, inform it that our fault
handler has been removed.
(myfault::faulted): Likewise for when the fault handler is installed
Brian Dessent wrote:
#define _CYGWIN_SIGNAL_STRING cYgSiGw00f
+#define _CYGWIN_FAULT_IGNORE_STRING cYgfAuLtIg
+#define _CYGWIN_FAULT_NOIGNORE_STRING cYgNofAuLtIg
Sigh, this breaks strace under Cygwin, I should have tested more. Sorry
about that. Apparently strace expects anything starting
Christopher Faylor wrote:
Thanks for the patch but I've been working on this too and, so far, I think
it is possible to have a very minimal way of dealing with this problem. I
haven't had time to delve into it too deeply but I have been exploring this
problem on and off for a couple of
Dave Korn wrote:
I'm having a conceptual difficulty here: Under what circumstances would you
expect there *not* to be a debugger attached to the
inferior to which the debugger is attached? That's a bit zen, isn't it?
The code in question here runs many times in the normal course of any
Gary Zablackis wrote:
I did some more research over the weekend and my
problem seems to only come when loading a dll via
dlopen() and run_ctors() is called from dll:init() and
pthread_key_create() is called during the time that
run_ctors() is active. I still have not found who is
calling
Corinna Vinschen wrote:
I must admit that I don't quite understand why that happens, but
when I save your patch into a file, all '=' characters are converted
into a '=3D' sequence. This is a bit weird given that you're using
us-ascii encoding. Does anybody know why this happens?
That's
in ?? ()
Brian2007-04-18 Brian Dessent [EMAIL PROTECTED]
* cygwin.sc: Remove duplicated .debug_macinfo section.
* dllfixdbg: Also copy DWARF-2 sections into .dbg file.
Index: cygwin.sc
===
RCS file: /cvs/src/src/winsup
Christopher Faylor wrote:
Thanks for doing this. Please check in. Can we switch to dwarf-2 by
default in the cygwin makefile(s)?
I thought about that, but the problem is anything you do to *FLAGS in
winsup/cygwin won't affect flags used in the other dirs like libiberty
or newlib, and so
Christopher Faylor wrote:
So, maybe a top-level configure option would be useful? At the very least
we can get rid of the -gstabs specific use in configure.
Oh, I didn't know there was anything that specifically mentions -gstabs
in there, just that if you don't set {C,CXX}FLAGS yourself
Corinna Vinschen wrote:
I debugged Cygwin native GDB a couple of days ago with code created by
gcc 4.2. It turned out that the DWARF2_UNWIND_INFO define set to 0
resulted in the DW_CFA_offset column missing. The result is that GDB is
unable to get the return address on the stack when using
Christopher Faylor wrote:
Let me rephrase the problem:
cygpath does not properly deal with the current directory
Thanks for the patch but we won't be applying it in this form.
I've been meaning to take a look at fixing this myself, because I'm
tired of:
$ cd /usr/bin
$ cygcheck ./ls
which fixes
the build for me.
Brian2007-06-16 Brian Dessent [EMAIL PROTECTED]
* include/sys/cdefs.h (__FBSDID): Define.
Index: include/sys/cdefs.h
===
RCS file: /cvs/src/src/winsup/cygwin/include/sys/cdefs.h,v
retrieving
/doc/Cygwin/rxvt-lt;vergt;.README/literal.
Unless my grep-fu failed that's it.
Brian2007-07-18 Brian Dessent [EMAIL PROTECTED]
* faq-using.xml (faq.using.console-window): Mention FHS location of
docs and remove outdated reference to ash.
Index: faq-using.xml
type as that generates a
compiler error, so I use an 'if'.
Brian2007-09-06 Brian Dessent [EMAIL PROTECTED]
* include/sys/stdio.h (_flockfile): Don't try to lock a FILE
that has the __SSTR flag set.
(_ftrylockfile): Likewise.
(_funlockfile): Likewise.
Index
Christopher Faylor wrote:
Go ahead and check this in but could you add a comment indicating that
this part of include/sys/stdio.h has to be kept in sync with newlib?
Done.
Nice catch!
I wish I could say I caught this by inspection but it was only by single
stepping through python guts that
structure) into newlib. The machinery to
compute these offsets already exists for the sake of gendef so
we might as well just use it here. */
Brian2007-09-06 Brian Dessent [EMAIL PROTECTED]
* libc/reent/getreent.c: Allow for case where __getreent is
defined as a macro.
Index
the whole 16MB file one byte at a
time, where this _REENT overhead would be most pronounced.
So, valid optimization or just complication?
Brian2007-09-06 Brian Dessent [EMAIL PROTECTED]
* include/cygwin/config.h (__getreent): Define inline version.
Index: include/cygwin/config.h
Brian Dessent wrote:
Done. I added the following comment to config.h to hopefully clarify
the situation:
/* The following provides an inline version of __getreent() for newlib,
which will be used throughout the library whereever there is a _r
version of a function that takes _REENT
Pedro Alves wrote:
The dllfixdbg hunk looks hard to read. Here's what is looks
like after patching:
I think that if whatever bugs used to exist in older binutils PE support
that necessitated this hackery are now gone, we can just do away with
dllfixdbg alltogether and just put this:
Pedro Alves wrote:
10 .cygheap 000a 611e 611e 2**2
ALLOC
11 .gnu_debuglink 0010 6128 6128 001d0a00 2**2
CONTENTS, READONLY, DEBUGGING
I'll come up with a different fix.
Just thinking out loud here... what
Brian Dessent wrote:
Fortunately, I have VMware with a Win98 image here.
The problem is that bloda.c calls NtQuerySystemInformation without using
any kind of autoload.cc-type indirection, and so cygcheck gets a hard
dependency on ntdll.dll which doesn't exist on 9x/ME. We need to do one
Brian Dessent wrote:
... but sadly a similar cleanup is
still required for cygpath as well if we are to support 9x/ME out of the
1.5 branch. In that case I suppose you could just revert cygpath.cc to
an older revision before the native APIs were added.
Er, nevermind. I was accidently
Christopher Faylor wrote:
Unless Corinna says differently, I think she wants to be in control of
what goes into the branch so I don't want to suggest that you should
check it in there too.
Okay, I'll let her take care of the branch since she's been handling all
the releases from it.
The
for cygwin/ and one in cygserver/.
Brian2008-02-03 Brian Dessent [EMAIL PROTECTED]
* smallprint.cc (__small_vsprintf): Use HEAP_NOTHEAP for type.
* strfuncs.cc (sys_wcstombs_alloc): Guard use of ccalloc
to !__OUTSIDE_CYGWIN__ for use in cygserver.
(sys_mbstowcs_alloc
Noel Burton-Krahn wrote:
The problem is there are several installable apps built on Cygwin,
like EAC, ClamAV, and one I just found which is a
Cygwin-on-a-thumbdrive. The problem is they can't all coexist because
they're distributed with different versions of the cygwin dlls.
Making them
I noticed in strace some lines like:
fhandler_base::close: closing
'/Device/NamedPipe/Win32Pipes.08e0.0002several junk bytes'
handle 0x740
This was caused by handle_to_fn simply forgetting to add a \0 when
converting, as in the attached patch.
Brian2008-03-07 Brian Dessent [EMAIL
necessary. I have tested builds in
both a combined tree as well as just winsup.
Brian2008-03-08 Brian Dessent [EMAIL PROTECTED]
* Makefile.in: Reorganize considerably, using GNU make's
static pattern rules and target-specific variables.
Makefile.in | 229
case. If you run make check it
builds the required files and runs the testsuite. I've attached a
sample output.
Brian2008-03-08 Brian Dessent [EMAIL PROTECTED]
* Makefile.in: Add a 'check' target that builds and runs
testsuite.exe from path-testsuite.o and testsuite.o
it to a differently-named object, but doing it otherwise would require
repeating the compile rule with all its ugly VERBOSE casing and I just
went to the trouble to eliminate all such repetition in the Makefile.
Brian2008-03-08 Brian Dessent [EMAIL PROTECTED]
* Makefile.in: Add a 'check
target in
find_app_on_path, we should temporarily set the CWD to the same dir as
the link, otherwise relative links will get normalized relative to
whatever dir cygcheck was run from.
Brian2008-03-08 Brian Dessent [EMAIL PROTECTED]
* cygcheck.cc (save_cwd_helper): New helper function
Corinna Vinschen wrote:
Doesn't that install testsuite.exe at `make install' time?
Ack, how about the attached?
Brian2008-03-09 Brian Dessent [EMAIL PROTECTED]
* Makefile.in (install): Don't install the testsuite.
Index: Makefile.in
Corinna Vinschen wrote:
Given that Cygwin changes to support long path names, I don't really
like to see new code still using MAX_PATH and Win32 Ansi functions
in the utils dir.
Regardless of this patch, path.cc:rel_vconcat() currently uses
GetCurrentDirectory() to resolve relative paths. It
Corinna Vinschen wrote:
I'm wondering if you would like to tweak the readlink functions, too.
Cygwin shortcuts can now have the path name appended to the actual
shortcut data. This hack was necessary to support pathnames longer than
2000 chars. See the comment and code in
Corinna Vinschen wrote:
Now that you mention it... did you see the comment in path.cc, line 3112ff?
There's a good chance that Windows shortcuts are not capable of long path
names. I didn't test it so far, but it would be certainly better for
readlink to use the POSIX path in the symlink
Christopher Faylor wrote:
I guess I misunderstood. I thought that the current working directory
could be derived through some complicated combination of Nt*() calls.
I could be wrong here but the way I understood it, there is no concept
of a working directory at the NT level, that is
Corinna Vinschen wrote:
The problem is that the cwd is stored as UNICODE_STRING with a
statically allocated buffer in the user parameter block. The
MaximumLength is set to 520. SetCurrentDirectory refuses to take paths
longer than that. In theory it would be possible to define a longer cwd
Corinna Vinschen wrote:
Btw., you don't need to make the buffers MAX_PATH + 1. MAX_PATH is
defined including the trailing NUL. Existing code shows a lot of
irritation about this...
Oh, I wasn't even thinking of that... the reason I used MAX_PATH + 1 was
because earlier I had written
+
Corinna Vinschen wrote:
Urgh. MAX_PATH is defined with trailing 0, SYMLINK_MAX is defined
without trailing 0 (like NAME_MAX). You should better change the
SYMLINK_MAX stuff back, afaics...
D'oh! 'Kay.
too
extreme.
Brian2008-03-13 Brian Dessent [EMAIL PROTECTED]
* ntdll.h: Add several missing NTSTATUS defines.
* pinfo.cc (pinfo::maybe_set_exit_code_from_windows): Recognise
and preserve the exit status of a child that terminated with
an unhandled exception
Eric Blake wrote:
Should we also mention 'cygcheck ./dll_not_found' to find out which ones
are missing?
It might be a good idea. On the other hand it's kind of long already.
I'm totally not married to what I've got for the wording though,
consider it a very rough draft.
|
Brian Dessent wrote:
isn't present, etc. I was really hoping to figure out a cool way to get
that info, perhaps by poking around in the TEB or PEB somewhere, but I
haven't gotten that far. If anyone has any general ideas where to look
for NTLDR's internal state, I'm all ears. I have
Christopher Faylor wrote:
That was going to be my first observation, actually. I'm still trying
to digest the patch but it seems like it wouldn't work well with the
fork retry code.
The patch doesn't change any behavior though: in current Cygwin if the
thing we're exec()ing returns a Win32
ReadFile, SetFilePointer, and GetFileInformationByHandle -- the
HANDLE is already open so this should require no change to support
WCHAR.
Brian2008-03-15 Brian Dessent [EMAIL PROTECTED]
* path.cc: Include malloc.h for alloca.
(is_symlink): Rewrite. Just read the whole file
paths, but really a reduced testcase is simply
open(c:/cygwin/bin/ls.exe, O_RDONLY) = ENOENT.
Brian2008-03-16 Brian Dessent [EMAIL PROTECTED]
* path.cc (normalize_posix_path): Correctly identify a win32
path beginning with a drive letter and forward slashes.
path.cc |2
Corinna Vinschen wrote:
len is a const value. Checking len for being 65536 is a constant
expression which always results in qddlen being 65535 so the ?: is
a noop, more or less.
Yeah, I realized that, and the compiler should optimize it away
completely. I put it explicitly as a test in the
Corinna Vinschen wrote:
Actually that was intended, but unfortunately the current path handling
deliberately creates DOS paths with slashes (in find_exec) right now,
so that doesn't work ATM.
I guess what I don't understand is how it's both possible for
open(c:/foo/bar.exe) to succeed and for
the correct location on the tls stack where
the real return address is stored.
Of course the labeling works for any module/dll, not just cygwin1.dll,
but I didn't have a more elaborate testcase to demonstrate.
Brian2008-03-18 Brian Dessent [EMAIL PROTECTED]
* exceptions.cc
Brian Dessent wrote:
Of course the labeling works for any module/dll, not just cygwin1.dll,
but I didn't have a more elaborate testcase to demonstrate.
Forgot to mention...
The symbols are just tacked on on the right hand side there for now. I
wasn't really sure how to handle that. I
Igor Peshansky wrote:
Would it make sense to force a newline before the function name and to
display it with a small indent? That way people who want the old-style
stackdump could just feed the new one into grep -v '^ ' or something...
Yes, that would be one way. That actually reminds me
Christopher Faylor wrote:
Sorry, but I don't like this concept. This bloats the cygwin DLL for a
condition that would be better served by either using gdb or generating
a real coredump.
I hear you, but part of the motivation for writing this was a recent
thread the other week on the gdb list
Corinna Vinschen wrote:
Is it a big problem to fix addr2line to deal with .dbg files?
I like your idea to add names to the stackdump especially because of
addr2line's brokenness. But, actually, if addr2line would work with
.dbg files, there would be no reason to add this to the stackdump
Brian Dessent wrote:
I think I see what's going on here though, the Cygwin fault handler took
the first chance exception and wrote the stackdump file, and only then
passed it on to the debugger, so that by the time gdb got notice of the
fault the stack was all fubar. This could be the reason
Brian Dessent wrote:
Yes, it means there is one frame that says sigbe instead of the actual
return location somewhere else. I don't think that's impossible to fix
either: the fault handler gets the context of the faulting thread so it
can look up its tls area through %fs:4 and peek
Christopher Faylor wrote:
After poking at this a little, I think it would be better to issue a
linux-like error message.
In my sandbox, I now have this:
bash-3.2$ ./libtest
/cygdrive/s/test/libtest.exe: error while loading shared libraries: liba.dll:
cannot open shared object file: No
-markup.
Unfortuantely there seems to also be some bitrot in the gprof side, as
the codepath to read BSD style gmon.out files is also broken. I've
posted a separate patch to the binutils list. With both these fixes,
gprof again works with Cygwin.
Brian2008-08-04 Brian Dessent [EMAIL PROTECTED
Christopher Faylor wrote:
Please use your best judgement about the +r/=r thing given Dave's
comments.
I think Dave's right, because when I compile with +r I get a may be
used uninitialized warning, so I'll just leave it as it is.
Both patches are now committed. I wonder how long this was
78 matches
Mail list logo