- Wine-devel - list.winehq.org

ReadFile and other issues
by eric pouech 21 Aug '01

21 Aug '01

while we're at talking about the ReadFile implementation with overlapped data, there's (an unrelated) point about ReadFile I'd like to discuss I've started looking at implementing proper console input in Wine (mainly the line mode editing). Apart from providing the functionnality, it would also allow to remove this unneeded code from the debugger the Win32 console input API is made of two layers 1/ the low level : each input event is stored into an INPUT_RECORD struct 2/ the high level which uses the low level events to make strings of them 1/ two functions are used: ReadConsoleInput(A|W) 2/ three functions are used: ReadFile and ReadConsole(A|W) the current implementation in Wine, makes the handle to the console input a file handle to the fd gotten from the console hence, the current implementation is broken in several ways - ReadFile is implemented as a raw read on the input fd, whereas it should be a higher level of function (forwarded from ReadFile to ReadConsole) - wait operations on handle are wrong: currently the handle is signaled as soon as a char is read (at the unix level) from the stdin fd, whereas it should be done at the level of the INPUT_RECORD queue to implement properly this (mainly the ability to forward ReadFile calls to ReadConsole for console handles), I'd like to do the following: - extend the get_handle_fd to also return an id corresponding to the type fd being used. As of today, handle / fd relationship mainly assumes that we can map a Windows operation on a handle onto a similar operation on a unix fd. The ReadFile/ReadConsoleInput doesn't fit in this category. - this type of feature would help in other areas too. For example, we never check that Comm functions really get a Comm handle... would such a mechanism be acceptable ? if so, should we do for every handle in the server, or just the ones we might convert into unix fd (file, consoles, pipes, sockets, "serials"...) A+ -- --------------- Eric Pouech (http://perso.wanadoo.fr/eric.pouech/) "The future will be better tomorrow", Vice President Dan Quayle

1 0

Re: icons uncoloured
by Gerard Patel 21 Aug '01

21 Aug '01

At 01:39 PM 21/08/2001 GMT, you wrote: >I've just done a cvs update and recompiled and now the icons come up in >gray, if at all. > >Is this a known issue? I suspect it could be solved by this one I just sent to wine-patches Gerard Index: graphics/x11drv/bitmap.c =================================================================== RCS file: /home/wine/wine/graphics/x11drv/bitmap.c,v retrieving revision 1.32 diff -u -r1.32 bitmap.c --- graphics/x11drv/bitmap.c 2001/08/20 19:19:18 1.32 +++ graphics/x11drv/bitmap.c 2001/08/21 14:17:07 @@ -73,7 +73,7 @@ if (hbitmap == dc->hBitmap) return hbitmap; /* nothing to do */ if (!(bmp = GDI_GetObjPtr( hbitmap, BITMAP_MAGIC ))) return 0; - if (bmp->header.dwCount) + if (bmp->header.dwCount && (hbitmap != GetStockObject(DEFAULT_BITMAP))) { WARN( "Bitmap already selected in another DC\n" ); GDI_ReleaseObj( hbitmap ); ~ ~

2 1

icons uncoloured
by Bill Medland 21 Aug '01

21 Aug '01

I've just done a cvs update and recompiled and now the icons come up in gray, if at all. Is this a known issue? Bill

1 0

Re: documentation make fix
by Francois Gouget 21 Aug '01

21 Aug '01

On Tue, 21 Aug 2001, Mike McCormack wrote: > > Not that anybody uses the documentation anyway... :-) > > ChangeLog: > * recurse all directories when making docs. > Don't fail if C_SRCS is empty. Maybe the attached patch would be a better way to do this: - use the regular make trick for recursing - test whether we have targets before entering the for loop Note that I did not really test it. Well, I have done a little testing now. And c2man is not happy if you don't create documentation/man3w beforehand... So I added an mkdir too, same for doc-html. -- Francois Gouget fgouget(a)free.fr http://fgouget.free.fr/ "Utilisateur" (nom commun) : Mot utilisé par les informaticiens en lieu et place d'"idiot".

1 0

Re: fix for Lawson's CLOCAL problem
by Mike McCormack 20 Aug '01

20 Aug '01

Hi Alexandre, >Of course, I have no problem with that. The create_file request takes >attributes, so create_serial can do the same (it would be better if >create_serial were done inside create_file, but that's another >story...) agreed. >> Another question: do you think it would be possible to implement a >> "write memory" message from the wineserver to the client? .. >But that would require multiple copies of the data (and many context >switches for large reads); that wouldn't be very efficient. I think >the current approach is much better. What is the problem with doing >CancelIo? IMO it would be messy to implement CancelIo with the current architecture. If it were all in the server (including actual read and write calls) it might be simpler... but probably not efficient, like you say. i think most of the data transferred from a serial port is pretty small though... but if the scheme was expanded to handle overlapped socket operations that wouldn't be true. Additionally, i though it might be useful to have a way to write small chunks of data to the client's memory. For example, when implementing WaitForNamedPipe and ConnectNamedPipe, i need to write a result code into an overlapped structure... it's only 4 bytes, but i need to create a new type of APC to deal with it. If i had a method to write directly into client memory, it would be simpler. Mike ------------------------------------------ mailto:Mike_McCormack(a)start.com.au ph +82 16 430 0425 __________________________________________________________________ Get your free Australian email account at http://www.Looksmart.com.au

1 0

Re: fix for Lawson's CLOCAL problem
by Alexandre Julliard 20 Aug '01

20 Aug '01

Mike McCormack <mike_mccormack(a)start.com.au> writes: > Hmmm. i'm not exactly sure about this. i don't think the server ever > blocked on serial port reads or opens... i think the reason Lawson had > problems was that the client thread was blocking in the overlapped > read code. (FILE_AsyncReadService in files/file.c) I don't know if this was Lawson problem, but open() can potentially block on a serial port, and we cannot have that in the server; so your change to set O_NONBLOCK is good, but it shouldn't depend on FILE_FLAG_OVERLAPPED. > If we use non-blocking fds unconditionally, a non overlapped ReadFile > may result in a busy loop in the client thread. Sure, you need to reset the O_NONBLOCK flag to do blocking reads. > Additionally, overlapped read (and write) requests should fail unless > the handle is opened with FILE_FLAG_OVERLAPPED. There are programs > that depend on that behaviour (as reported by Eric Kohl, 3 Feb 2001 on > this list). This is another reason to propagate CreateFile's > attributes parameter into the wineserver. Of course, I have no problem with that. The create_file request takes attributes, so create_serial can do the same (it would be better if create_serial were done inside create_file, but that's another story...) > Another question: do you think it would be possible to implement a > "write memory" message from the wineserver to the client? This message > would be carried over wait_fd and would be received by the client > asynchronously in wait_reply(). The message would contain a length, > address and data to be written. If this is possible, i could move > async i/o back into the server, keep it efficient and perhaps > implement CancelIo properly... But that would require multiple copies of the data (and many context switches for large reads); that wouldn't be very efficient. I think the current approach is much better. What is the problem with doing CancelIo? -- Alexandre Julliard julliard(a)winehq.com

1 0

Re: winsock stuff
by Daniel Walker 20 Aug '01

20 Aug '01

This seems to work.. Daniel Walker --- server/sock.c 2001/05/14 20:09:39 1.18 +++ server/sock.c 2001/08/21 01:31:04 @@ -113,9 +113,14 @@ { struct sock *sock = (struct sock *)obj; unsigned int emask; + char rBuf; assert( sock->obj.ops == &sock_ops ); if (debug_level) fprintf(stderr, "socket %d select event: %x\n", sock->obj.fd, event); + + if (event == POLLIN && !recvfrom(sock->obj.fd, &rBuf, sizeof(rBuf), MSG_PEEK, NULL, NULL)) + event = POLLHUP; + if (sock->state & WS_FD_CONNECT) { /* connecting */

1 0

Re: fix for Lawson's CLOCAL problem
by Mike McCormack 20 Aug '01

20 Aug '01

Hi Alexandre, Hmmm. i'm not exactly sure about this. i don't think the server ever blocked on serial port reads or opens... i think the reason Lawson had problems was that the client thread was blocking in the overlapped read code. (FILE_AsyncReadService in files/file.c) If we use non-blocking fds unconditionally, a non overlapped ReadFile may result in a busy loop in the client thread. Additionally, overlapped read (and write) requests should fail unless the handle is opened with FILE_FLAG_OVERLAPPED. There are programs that depend on that behaviour (as reported by Eric Kohl, 3 Feb 2001 on this list). This is another reason to propagate CreateFile's attributes parameter into the wineserver. Another question: do you think it would be possible to implement a "write memory" message from the wineserver to the client? This message would be carried over wait_fd and would be received by the client asynchronously in wait_reply(). The message would contain a length, address and data to be written. If this is possible, i could move async i/o back into the server, keep it efficient and perhaps implement CancelIo properly... Mike > The open on the server side must always be non-blocking, we cannot > have the server block on anything. If there are cases where we need to > block on open, it will have to be done on the client side. ------------------------------------------ mailto:Mike_McCormack(a)start.com.au ph +82 16 430 0425 __________________________________________________________________ Get your free Australian email account at http://www.Looksmart.com.au

1 0

Re: winsock stuff
by Alexandre Julliard 20 Aug '01

20 Aug '01

Daniel Walker <diwalker(a)earthlink.net> writes: > How is it unreliable? Why is MSG_PEEK more reliable? I'm not opposed to > using MSG_PEEK , just curios.. FIONREAD is unreliable because you cannot distinguish between an EOF condition and an empty receive buffer. recvmsg(MSG_PEEK) is supposed to return 0 on EOF, and EWOULDBLOCK on empty buffer, so you can determine the real status. > Another issue is why doesn't poll() give > us a POLLHUP? That would be ideal, get poll() to work the way we need it > too. As I understand it poll() returns POLLHUP only when both directions of the socket have been closed; but if you shutdown() only the read or write side, you don't get POLLHUP so that you can continue using poll() for the other direction. This has apparently been changed between kernels 2.2 and 2.4; in 2.2 POLLHUP is returned as soon as the read side of the socket has been closed. -- Alexandre Julliard julliard(a)winehq.com

1 0

Re: winsock stuff
by Alexandre Julliard 20 Aug '01

20 Aug '01

Daniel Walker <diwalker(a)earthlink.net> writes: > To improve on this patch we could only do the FIONREAD if the sock is > in a non-blocking state. That's the only time that doing the FIONREAD > would be an advantage. Also we wouldn't need to worry about screwing up > blocking sockets. Even further, we could only do the FIONREAD if socket > is non-blocking and looking for an FD_CLOSE event. Using FIONREAD is simply not an option, it cannot be made reliable. A possibility to detect EOF could be a recvmsg with MSG_PEEK flag, though you should probably verify first that it behaves as expected. -- Alexandre Julliard julliard(a)winehq.com

2 1