You can subscribe to this list here.
2006 |
Jan
|
Feb
|
Mar
|
Apr
|
May
|
Jun
(2) |
Jul
(1) |
Aug
|
Sep
|
Oct
|
Nov
|
Dec
|
---|---|---|---|---|---|---|---|---|---|---|---|---|
2007 |
Jan
|
Feb
|
Mar
|
Apr
|
May
|
Jun
|
Jul
|
Aug
(1) |
Sep
|
Oct
|
Nov
|
Dec
|
2009 |
Jan
|
Feb
|
Mar
|
Apr
|
May
|
Jun
|
Jul
|
Aug
|
Sep
|
Oct
(2) |
Nov
(29) |
Dec
(13) |
2010 |
Jan
(5) |
Feb
(5) |
Mar
|
Apr
|
May
(4) |
Jun
|
Jul
(1) |
Aug
|
Sep
|
Oct
|
Nov
|
Dec
|
2011 |
Jan
|
Feb
|
Mar
|
Apr
|
May
(3) |
Jun
|
Jul
|
Aug
|
Sep
|
Oct
|
Nov
|
Dec
|
From: Tim N. <tim...@gm...> - 2011-05-09 22:58:00
|
ack, sorry, I misunderstood.. I thought these were for ninefs, not npfs. On Mon, May 9, 2011 at 7:28 AM, Tim Newsham <tim...@gm...> wrote: > thank you. I'll try to get those into the tree today. > > On Sun, May 8, 2011 at 7:03 PM, Mike Ryan <mik...@la...> wrote: >> Three things: >> >> 0) [enhancement] >> I wanted npfs to run in the background by default, but added in a >> flag for running it in the foreground. >> >> 1) [patch] >> stat(2) fails on large files unless _FILE_OFFSET_BITS is defined to >> be 64, so this patch enables that in the right Makefiles. >> >> 2) [patch] >> More subtle: when a file is owned by an unknown user or group, 9pfs >> was returning an error. I've modified it to return numeric UIDs and >> GIDs in that situation. >> >> ------------------------------------------------------------------------------ >> WhatsUp Gold - Download Free Network Management Software >> The most intuitive, comprehensive, and cost-effective network >> management toolset available today. Delivers lowest initial >> acquisition cost and overall TCO of any competing solution. >> http://p.sf.net/sfu/whatsupgold-sd >> _______________________________________________ >> Npfs-devel mailing list >> Npf...@li... >> https://lists.sourceforge.net/lists/listinfo/npfs-devel >> >> > > > > -- > Tim Newsham | www.thenewsh.com/~newsham | thenewsh.blogspot.com > -- Tim Newsham | www.thenewsh.com/~newsham | thenewsh.blogspot.com |
From: Tim N. <tim...@gm...> - 2011-05-09 17:29:02
|
thank you. I'll try to get those into the tree today. On Sun, May 8, 2011 at 7:03 PM, Mike Ryan <mik...@la...> wrote: > Three things: > > 0) [enhancement] > I wanted npfs to run in the background by default, but added in a > flag for running it in the foreground. > > 1) [patch] > stat(2) fails on large files unless _FILE_OFFSET_BITS is defined to > be 64, so this patch enables that in the right Makefiles. > > 2) [patch] > More subtle: when a file is owned by an unknown user or group, 9pfs > was returning an error. I've modified it to return numeric UIDs and > GIDs in that situation. > > ------------------------------------------------------------------------------ > WhatsUp Gold - Download Free Network Management Software > The most intuitive, comprehensive, and cost-effective network > management toolset available today. Delivers lowest initial > acquisition cost and overall TCO of any competing solution. > http://p.sf.net/sfu/whatsupgold-sd > _______________________________________________ > Npfs-devel mailing list > Npf...@li... > https://lists.sourceforge.net/lists/listinfo/npfs-devel > > -- Tim Newsham | www.thenewsh.com/~newsham | thenewsh.blogspot.com |
From: Mike R. <mik...@la...> - 2011-05-09 05:03:38
|
Three things: 0) [enhancement] I wanted npfs to run in the background by default, but added in a flag for running it in the foreground. 1) [patch] stat(2) fails on large files unless _FILE_OFFSET_BITS is defined to be 64, so this patch enables that in the right Makefiles. 2) [patch] More subtle: when a file is owned by an unknown user or group, 9pfs was returning an error. I've modified it to return numeric UIDs and GIDs in that situation. |
From: Latchesar I. <lu...@io...> - 2010-07-02 17:08:44
|
Hi, Sorry for the late response. I am not sure how useful this change is. As you mention, it would work only for specific file servers. If you connect to non-.u server, you better check the text message instead of depending on ecode. Thanks, Lucho On Sat, May 29, 2010 at 6:12 AM, Georg Lehner <jor...@ma...> wrote: > Hello! > > find attached a patch for npfs, adding error-string to error code > handling for some 9p error messages from the server. > > in libnpfs/np.c the function np_errno() provides the following heuristics: > > - "file does not exist" -> ENOENT > - "permission denied" -> EACCES > - "file already exists" -> EEXIST > - "is a directory" -> EACCES > > np_errno() is used in np_deserialize() in the non-dotu branch instead of > returning ~0. > > This simple approach requires changeing np_errno() for adding error > messages. It has been tested only with Plan9 fossil file server. > > Regards, > > Jorge-León > > > > > ------------------------------------------------------------------------------ > > > _______________________________________________ > Npfs-devel mailing list > Npf...@li... > https://lists.sourceforge.net/lists/listinfo/npfs-devel > > |
From: Georg L. <jor...@ma...> - 2010-05-29 14:42:15
|
Hello! First please apologize for the misleading subject of the previous post. instead of "Patch for mount with attach name" it should have been "Patch for additional errorhandling" Find attached in the current message a patch for npfs, adding setting of the errorstring at several places where it is not done yet. In libnpclient/{open,walk}.c and libnpfs/np.c several function return with error (NULL), but without setting the errorstring. This leads to ambiguous return conditions on error. With this patch, np_werror() has been inserted before returning. In libnpclient/walk.c there is also a part, where additional errorhandling is done. After calling npc_fid_alloc() the return value is checked for beeing not NULL (right now I see, that at this spot i did not call np_werror(), please consider adding if required). Regards, Jorge-León |
From: Georg L. <jor...@ma...> - 2010-05-29 12:47:29
|
Hello! find attached a patch for npfs, adding additional errorhandling at some spots in libnpclient/{open,walk}.c. Essentially the return value of np_create_*() calls is checked for beeing not null before using it in npc_rpc() calls. While it should not hurt, i have not the insight into the code to be able to see if this is required at all points where i added the checks, so please review. Regards, Jorge-León |
From: Georg L. <jor...@ma...> - 2010-05-29 12:39:21
|
Hello! find attached a patch for npfs, adding error-string to error code handling for some 9p error messages from the server. in libnpfs/np.c the function np_errno() provides the following heuristics: - "file does not exist" -> ENOENT - "permission denied" -> EACCES - "file already exists" -> EEXIST - "is a directory" -> EACCES np_errno() is used in np_deserialize() in the non-dotu branch instead of returning ~0. This simple approach requires changeing np_errno() for adding error messages. It has been tested only with Plan9 fossil file server. Regards, Jorge-León |
From: Georg L. <jor...@ma...> - 2010-05-21 21:07:14
|
Hello! find attached a patch for npfs, adding a function for mounting a different file system then the default one. The new function, npfs_netmounta() is directly derived from the original npfs_netmount() function, by adding a parameter for the attach name. npfs_netmount() now just is a wrapper for npfs_netmounta(), with the attach name parameter set to NULL. This can be used for example with ninefs to map one Windows drive to the active file system of a fossil/venti server and another one to the dump file system of the same server. Regards, Jorge-León |
From: Latchesar I. <lu...@io...> - 2010-02-16 04:03:40
|
If path has more than MAXWELEM it will be valid and has to be clunked. Thanks, Lucho On Mon, Feb 15, 2010 at 6:26 PM, Tim Newsham <tim...@gm...> wrote: > http://plan9.bell-labs.com/magic/man2html/5/walk says that the fid is only > valid if the > walk returns an array of qid's that is the same length as the array of path > components > you passed in. It doesn't return an error (Rerror) but the fid is still not > valid. You'll > see in traces talking to plan9 fileservers that if you use npc_walk (ie. > from npc_open) > on a file that doesnt exist, it will walk most of the way, return success, > then do a > Tclunk which will fail saying that the fid doesn't exist. > > On Mon, Feb 15, 2010 at 3:01 PM, Latchesar Ionkov <lu...@io...> wrote: >> >> Ack for the first patch. >> >> The fid needs to be clunked if there is a partial success while >> walking. For example, if Twalk tries to walk four wnames and only >> three succeed, the fid is valid and should be clunked because we don't >> return it when we exit from the function. >> >> Thanks, >> Lucho >> >> On Mon, Feb 15, 2010 at 1:50 PM, Tim Newsham <tim...@gm...> >> wrote: >> > Two patches attached for approval: >> > - emptystat.patch adds the unspecified value for the group. I forgot >> > to >> > add it earlier >> > - walkerr.patch removes the clunk that walk does when an error >> > happens. >> > All paths to error happen when no new fid was opened so the clunk >> > is >> > extraneous and only serves to ruin the stored error code. >> > (Note: fid is only opened when the walk completely walks all path >> > components). >> > >> > -- >> > Tim Newsham | www.thenewsh.com/~newsham | thenewsh.blogspot.com >> > >> > >> > >> > ------------------------------------------------------------------------------ >> > SOLARIS 10 is the OS for Data Centers - provides features such as >> > DTrace, >> > Predictive Self Healing and Award Winning ZFS. Get Solaris 10 NOW >> > http://p.sf.net/sfu/solaris-dev2dev >> > _______________________________________________ >> > Npfs-devel mailing list >> > Npf...@li... >> > https://lists.sourceforge.net/lists/listinfo/npfs-devel >> > >> > > > > > -- > Tim Newsham | www.thenewsh.com/~newsham | thenewsh.blogspot.com > > |
From: Tim N. <tim...@gm...> - 2010-02-16 01:26:57
|
http://plan9.bell-labs.com/magic/man2html/5/walk says that the fid is only valid if the walk returns an array of qid's that is the same length as the array of path components you passed in. It doesn't return an error (Rerror) but the fid is still not valid. You'll see in traces talking to plan9 fileservers that if you use npc_walk (ie. from npc_open) on a file that doesnt exist, it will walk most of the way, return success, then do a Tclunk which will fail saying that the fid doesn't exist. On Mon, Feb 15, 2010 at 3:01 PM, Latchesar Ionkov <lu...@io...> wrote: > Ack for the first patch. > > The fid needs to be clunked if there is a partial success while > walking. For example, if Twalk tries to walk four wnames and only > three succeed, the fid is valid and should be clunked because we don't > return it when we exit from the function. > > Thanks, > Lucho > > On Mon, Feb 15, 2010 at 1:50 PM, Tim Newsham <tim...@gm...> > wrote: > > Two patches attached for approval: > > - emptystat.patch adds the unspecified value for the group. I forgot > to > > add it earlier > > - walkerr.patch removes the clunk that walk does when an error > happens. > > All paths to error happen when no new fid was opened so the clunk is > > extraneous and only serves to ruin the stored error code. > > (Note: fid is only opened when the walk completely walks all path > > components). > > > > -- > > Tim Newsham | www.thenewsh.com/~newsham | thenewsh.blogspot.com > > > > > > > ------------------------------------------------------------------------------ > > SOLARIS 10 is the OS for Data Centers - provides features such as DTrace, > > Predictive Self Healing and Award Winning ZFS. Get Solaris 10 NOW > > http://p.sf.net/sfu/solaris-dev2dev > > _______________________________________________ > > Npfs-devel mailing list > > Npf...@li... > > https://lists.sourceforge.net/lists/listinfo/npfs-devel > > > > > -- Tim Newsham | www.thenewsh.com/~newsham | thenewsh.blogspot.com |
From: Latchesar I. <lu...@io...> - 2010-02-16 01:01:17
|
Ack for the first patch. The fid needs to be clunked if there is a partial success while walking. For example, if Twalk tries to walk four wnames and only three succeed, the fid is valid and should be clunked because we don't return it when we exit from the function. Thanks, Lucho On Mon, Feb 15, 2010 at 1:50 PM, Tim Newsham <tim...@gm...> wrote: > Two patches attached for approval: > - emptystat.patch adds the unspecified value for the group. I forgot to > add it earlier > - walkerr.patch removes the clunk that walk does when an error happens. > All paths to error happen when no new fid was opened so the clunk is > extraneous and only serves to ruin the stored error code. > (Note: fid is only opened when the walk completely walks all path > components). > > -- > Tim Newsham | www.thenewsh.com/~newsham | thenewsh.blogspot.com > > > ------------------------------------------------------------------------------ > SOLARIS 10 is the OS for Data Centers - provides features such as DTrace, > Predictive Self Healing and Award Winning ZFS. Get Solaris 10 NOW > http://p.sf.net/sfu/solaris-dev2dev > _______________________________________________ > Npfs-devel mailing list > Npf...@li... > https://lists.sourceforge.net/lists/listinfo/npfs-devel > > |
From: Latchesar I. <lu...@io...> - 2010-01-05 20:10:58
|
I am fine with simpleusers being included in the POSIX library. Acked-by: Latchesar Ionkov <lu...@io...> On Tue, Jan 5, 2010 at 12:16 PM, Tim Newsham <tim...@gm...> wrote: > Uxusers was not properly failing when looking up usernames that > dont exist on the local system. The attached patch fixes this. > It occurs to me now that it would be useful for server writers to be > able to pick uxusers or simpleusers for their userpool depending > on if they want to line up with local unix accounts or not. Currently > simpleusers is only used in win32 (I wrote it because win32 doesnt > support the same unix user api). I think it might be best for uxusers > to be the default when present, but still provide access to simpleusers > even on unix systems. > > -- > Tim Newsham | www.thenewsh.com/~newsham | thenewsh.blogspot.com > > > ------------------------------------------------------------------------------ > This SF.Net email is sponsored by the Verizon Developer Community > Take advantage of Verizon's best-in-class app development support > A streamlined, 14 day to market process makes app distribution fast and easy > Join now and get one step closer to millions of Verizon customers > http://p.sf.net/sfu/verizon-dev2dev > _______________________________________________ > Npfs-devel mailing list > Npf...@li... > https://lists.sourceforge.net/lists/listinfo/npfs-devel > > |
From: Tim N. <tim...@gm...> - 2010-01-05 19:39:11
|
I committed the libnpauth stuff. There are examples in subdirs of the library. Please let me know if anyone has problems with anything. -- Tim Newsham | www.thenewsh.com/~newsham | thenewsh.blogspot.com |
From: Latchesar I. <lu...@io...> - 2010-01-05 16:34:44
|
Sorry, forgot to respond. The auth library looks good. You are the Windows expert, so I can't say anything about the windows patch. :) Ack for both. Thanks, Lucho On Mon, Jan 4, 2010 at 8:32 PM, Tim Newsham <tim...@gm...> wrote: > Anyone have a chance to review these yet? Shall I submit? Need more time > (on vacation)? > Happy New Years... > On Wed, Dec 30, 2009 at 2:49 PM, Tim Newsham <tim...@gm...> wrote: >> >> Two files attached. >> The first is libnpauth.tgz which has a new libnpauth directory for under >> the npfs/npfs/trunk directory. The npauth.h should be moved to the include >> directory, all other files belong in place. This library has client and >> server functions for authenticating using p9sk1 and p9any (supporting p9sk1 >> only). They've been tested to compile and run under Ubuntu, OSX and Win32. >> It includes two examples that test out the server and the client >> authentication, respectively. This is all fresh new code. Feel free to >> criticize anything including style, layout, functionality, API design. The >> goal is to make it fit well with the rest of npfs. >> The second file is a small patch for the winthread implementation in >> libnpfs to remove a debugging assert I forgot to remove earlier. >> >> I ended up ditching the idea of looking into the Npfsys in some way to >> figure out where the authentication server may be, in favor of just having >> the API client specify where to find the auth server. >> -- >> Tim Newsham | www.thenewsh.com/~newsham | thenewsh.blogspot.com >> > > > > -- > Tim Newsham | www.thenewsh.com/~newsham | thenewsh.blogspot.com > > > ------------------------------------------------------------------------------ > This SF.Net email is sponsored by the Verizon Developer Community > Take advantage of Verizon's best-in-class app development support > A streamlined, 14 day to market process makes app distribution fast and easy > Join now and get one step closer to millions of Verizon customers > http://p.sf.net/sfu/verizon-dev2dev > _______________________________________________ > Npfs-devel mailing list > Npf...@li... > https://lists.sourceforge.net/lists/listinfo/npfs-devel > > |
From: Tim N. <tim...@gm...> - 2010-01-05 03:32:12
|
Anyone have a chance to review these yet? Shall I submit? Need more time (on vacation)? Happy New Years... On Wed, Dec 30, 2009 at 2:49 PM, Tim Newsham <tim...@gm...> wrote: > Two files attached. > > The first is libnpauth.tgz which has a new libnpauth directory for under > the npfs/npfs/trunk directory. The npauth.h should be moved to the include > directory, all other files belong in place. This library has client and > server functions for authenticating using p9sk1 and p9any (supporting p9sk1 > only). They've been tested to compile and run under Ubuntu, OSX and Win32. > It includes two examples that test out the server and the client > authentication, respectively. This is all fresh new code. Feel free to > criticize anything including style, layout, functionality, API design. The > goal is to make it fit well with the rest of npfs. > > The second file is a small patch for the winthread implementation in > libnpfs to remove a debugging assert I forgot to remove earlier. > > I ended up ditching the idea of looking into the Npfsys in some way to > figure out where the authentication server may be, in favor of just having > the API client specify where to find the auth server. > > -- > Tim Newsham | www.thenewsh.com/~newsham | thenewsh.blogspot.com > > -- Tim Newsham | www.thenewsh.com/~newsham | thenewsh.blogspot.com |
From: Tim N. <tim...@gm...> - 2009-12-31 02:07:39
|
Right now my portable code has an #ifdef for win32 code to initialize the winsock library. Would it be better to create an np_init function that hides this platform difference and require all npfs programs to call np_init during startup? -- Tim Newsham | www.thenewsh.com/~newsham | thenewsh.blogspot.com |
From: Latchesar I. <lu...@io...> - 2009-12-29 23:23:41
|
Acked-by: Latchesar Ionkov <lu...@io...> On Tue, Dec 29, 2009 at 2:42 PM, Tim Newsham <tim...@gm...> wrote: > Attached are two small patches > - tiny mkfile fix for the npclient examples. OSX doesnt like space after > "-L" > - add a hint for getaddrinfo to force it to use SOCK_STREAM. Otherwise I > see it trying to use a UDP socket on several machines. Previously I was > hacking around this manually. > The getaddrinfo patch removes and ignores an ominous warning in the code. > I tested in OSX and Ubuntu8 and it seems to work fine in both places. I > don't > know if and where it might break, but I do know the current code is breaking > on several platforms. > > -- > Tim Newsham | www.thenewsh.com/~newsham | thenewsh.blogspot.com > > > ------------------------------------------------------------------------------ > This SF.Net email is sponsored by the Verizon Developer Community > Take advantage of Verizon's best-in-class app development support > A streamlined, 14 day to market process makes app distribution fast and easy > Join now and get one step closer to millions of Verizon customers > http://p.sf.net/sfu/verizon-dev2dev > _______________________________________________ > Npfs-devel mailing list > Npf...@li... > https://lists.sourceforge.net/lists/listinfo/npfs-devel > > |
From: Latchesar I. <lu...@io...> - 2009-12-29 16:16:24
|
Acked-by: Latchesar Ionkov <lu...@io...> On Mon, Dec 28, 2009 at 12:07 PM, Tim Newsham <tim...@gm...> wrote: > err.. correction: it specifies which version the *client* will use since > some *servers* do not recognized "9P2000.u" as valid. > > On Mon, Dec 28, 2009 at 8:48 AM, Tim Newsham <tim...@gm...> wrote: >> >> The attached patch changes the mount api in the libnpclient library >> to allow the library user to specify whether or not to support dotu. >> In the current code the server will accept always accept dotu >> connections. This can cause problems for some clients that do not >> recognize "9P2000.u" as a valid version string. >> The patch also adds a -U flag to the examples in the examples >> directory. >> -- >> Tim Newsham | www.thenewsh.com/~newsham | thenewsh.blogspot.com >> > > > > -- > Tim Newsham | www.thenewsh.com/~newsham | thenewsh.blogspot.com > > > ------------------------------------------------------------------------------ > This SF.Net email is sponsored by the Verizon Developer Community > Take advantage of Verizon's best-in-class app development support > A streamlined, 14 day to market process makes app distribution fast and easy > Join now and get one step closer to millions of Verizon customers > http://p.sf.net/sfu/verizon-dev2dev > _______________________________________________ > Npfs-devel mailing list > Npf...@li... > https://lists.sourceforge.net/lists/listinfo/npfs-devel > > |
From: Latchesar I. <lu...@io...> - 2009-12-29 16:15:55
|
A separate library sounds good. I don't like exposing the file descriptor. May be a better way of getting the remote address is to have a getter that returns it (if available). It probably makes sense to have the same on the server side. Thanks, Lucho On Thu, Dec 24, 2009 at 1:07 PM, Tim Newsham <tim...@gm...> wrote: > I've written some code to do client authentication using p9sk1 or p9any. > I'd be interested in committing it to npfs. The big downside is it adds > a depedency on openssl (for the crypto support). Three options I can > think of: > - dont put it in the tree, keep it as an external library. > - put it in the tree as a separate library (libnpauth?). Those > who dont want to build the auth library wont need openssl. > - put it in libnpclient. If you dont want openssl you'd have to > take it out of your makefile manually. This is where it fits > most naturally, build considerations aside. > thoughts? > also, if you prefer a different crypto library, I'm flexible. I just picked > openssl because its my default. > I'll be working on server-side auth next... > ps: I had to peek into the opaque Npcfsys structure. I did this to > get fsys->fd so I can getpeername and then connect to the same > host on the auth port. > -- > Tim Newsham | www.thenewsh.com/~newsham | thenewsh.blogspot.com > > > ------------------------------------------------------------------------------ > This SF.Net email is sponsored by the Verizon Developer Community > Take advantage of Verizon's best-in-class app development support > A streamlined, 14 day to market process makes app distribution fast and easy > Join now and get one step closer to millions of Verizon customers > http://p.sf.net/sfu/verizon-dev2dev > _______________________________________________ > Npfs-devel mailing list > Npf...@li... > https://lists.sourceforge.net/lists/listinfo/npfs-devel > > |
From: Tim N. <tim...@gm...> - 2009-12-28 19:08:00
|
err.. correction: it specifies which version the *client* will use since some *servers* do not recognized "9P2000.u" as valid. On Mon, Dec 28, 2009 at 8:48 AM, Tim Newsham <tim...@gm...> wrote: > The attached patch changes the mount api in the libnpclient library > to allow the library user to specify whether or not to support dotu. > In the current code the server will accept always accept dotu > connections. This can cause problems for some clients that do not > recognize "9P2000.u" as a valid version string. > The patch also adds a -U flag to the examples in the examples > directory. > > -- > Tim Newsham | www.thenewsh.com/~newsham | thenewsh.blogspot.com > > -- Tim Newsham | www.thenewsh.com/~newsham | thenewsh.blogspot.com |
From: Tim N. <tim...@gm...> - 2009-12-24 20:08:06
|
I've written some code to do client authentication using p9sk1 or p9any. I'd be interested in committing it to npfs. The big downside is it adds a depedency on openssl (for the crypto support). Three options I can think of: - dont put it in the tree, keep it as an external library. - put it in the tree as a separate library (libnpauth?). Those who dont want to build the auth library wont need openssl. - put it in libnpclient. If you dont want openssl you'd have to take it out of your makefile manually. This is where it fits most naturally, build considerations aside. thoughts? also, if you prefer a different crypto library, I'm flexible. I just picked openssl because its my default. I'll be working on server-side auth next... ps: I had to peek into the opaque Npcfsys structure. I did this to get fsys->fd so I can getpeername and then connect to the same host on the auth port. -- Tim Newsham | www.thenewsh.com/~newsham | thenewsh.blogspot.com |
From: Latchesar I. <lu...@io...> - 2009-12-22 03:22:56
|
I am not sure who wrote that code (I think it wasn't me) and what is the story behind the comment. Thanks, Lucho On Mon, Dec 21, 2009 at 7:59 PM, Tim Newsham <tim...@gm...> wrote: > In libnpclient/netmount.c: > /* they have this cute 'hints' thing you can put in. > * it would be really great if it worked, but it fails in some > * places, so just don't use it. > */ > r = getaddrinfo(name, port, NULL, &addrlist); > could you expound on what sort of errors you've seen with the hints? > I've run into a problem several times now where npc_netaddr ends > up returning a SOCK_DGRAM reference to UDP port 564 instead > of SOCK_STREAM for TCP port 564. I imagine this is what the > hint is supposed to avoid. I'm seeing it now in OS X 10.5.6 and > I saw it before (I cant remember if it was in ubuntu or in windows). > If I changed the code to use the hint, what would the harm be and > on what platform? > -- > Tim Newsham | www.thenewsh.com/~newsham | thenewsh.blogspot.com > > > ------------------------------------------------------------------------------ > This SF.Net email is sponsored by the Verizon Developer Community > Take advantage of Verizon's best-in-class app development support > A streamlined, 14 day to market process makes app distribution fast and easy > Join now and get one step closer to millions of Verizon customers > http://p.sf.net/sfu/verizon-dev2dev > _______________________________________________ > Npfs-devel mailing list > Npf...@li... > https://lists.sourceforge.net/lists/listinfo/npfs-devel > > |
From: Tim N. <tim...@gm...> - 2009-12-22 02:59:49
|
In libnpclient/netmount.c: /* they have this cute 'hints' thing you can put in. * it would be really great if it worked, but it fails in some * places, so just don't use it. */ r = getaddrinfo(name, port, NULL, &addrlist); could you expound on what sort of errors you've seen with the hints? I've run into a problem several times now where npc_netaddr ends up returning a SOCK_DGRAM reference to UDP port 564 instead of SOCK_STREAM for TCP port 564. I imagine this is what the hint is supposed to avoid. I'm seeing it now in OS X 10.5.6 and I saw it before (I cant remember if it was in ubuntu or in windows). If I changed the code to use the hint, what would the harm be and on what platform? -- Tim Newsham | www.thenewsh.com/~newsham | thenewsh.blogspot.com |
From: Abhishek K. <abb...@gm...> - 2009-12-07 05:08:46
|
since spc_netmount does not use inet_ntoa (because it is not thread-safe, uggh), it runs into endian issues when extracting the address from in_addr_t. xcpu2 looks for an invalid terminal server address on linux ppc64. this patch fixes the issue. Signed-off-by: Abhishek Kulkarni <adk...@um...> |
From: Latchesar I. <lu...@io...> - 2009-11-23 16:04:27
|
Acked-by: Latchesar Ionkov <lu...@io...> On Mon, Nov 23, 2009 at 8:47 AM, Abhishek Kulkarni <abb...@gm...> wrote: > this must have bit me so many times by now. > addrlist is NULL when getaddrinfo fails. so we > simply check before using it. > > using gai_strerror() to get the actual error string > would be good, but it doesn't look to be > thread-safe. > > Signed-off-by: Abhishek Kulkarni <adk...@um...> > > Index: libnpclient/netmount.c > =================================================================== > --- libnpclient/netmount.c (revision 48) > +++ libnpclient/netmount.c (working copy) > @@ -83,6 +83,9 @@ > int fd; > char ename[32]; > > + if (!addrlist) > + goto error; > + > fd = socket(addrlist->ai_family, addrlist->ai_socktype, 0); > if (fd < 0) { > np_uerror(errno); > > ------------------------------------------------------------------------------ > Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day > trial. Simplify your report design, integration and deployment - and focus on > what you do best, core application coding. Discover what's new with > Crystal Reports now. http://p.sf.net/sfu/bobj-july > _______________________________________________ > Npfs-devel mailing list > Npf...@li... > https://lists.sourceforge.net/lists/listinfo/npfs-devel > |