Re: Two different platforms sharing common code base: How to do so?
Spiro Trikaliotis wrote: Well, it depends upon the tool you're using. We are building drivers, and they need the build.exe tool, which is part of the DDK. That one can only compile files in the current directory. As John has pointed out, it might even be able to compile files in the parent directory, which would help here. I've done several developments for Windows drivers where some of the driver source was shared with Linux and BSD drivers. The solution we came up with was to write a perl build script for the Linux BSD drivers which generated a makefile (in a build directory) from the data accumulated from the sources files in each directory. The advantages of this are: 1. only have to modify one file when adding a new source file. 2. Linux drivers can be built using supported mechanism e.g. kbuild for linux. 3. Windows drivers are built in DDK environment using dirs/sources mechanism as supported by MS. So this means that the Windows Makefiles are stored in the main source tree and the Linux Makefile is generated by a script and is placed in a separate build directory. Mike ___ Info-cvs mailing list [EMAIL PROTECTED] http://lists.gnu.org/mailman/listinfo/info-cvs
Re: Ineroperability problem with Tortoise CVS
-BEGIN PGP SIGNED MESSAGE- Hash: SHA1 Not without hacking the source code. Okay fair enough. You may wish to file a bug report with the Tortoise CVS folks and/or CVSNT folks. Looks like its unneccessary. The log parsing code in Tortoise CVS has been rewritten since the last stable release and the new parser looks like it will handle the new date format without problems. I'll grab a test release to check and report a bug if there are still problems. For now, you will probably need to go in and hack the src/main.c::format_time_t() and src/main.c:gmformat_time_t() functions by hand. Thanks for that. If the end users of the system complain I know where to go and hack. Mike ___ Info-cvs mailing list [EMAIL PROTECTED] http://lists.gnu.org/mailman/listinfo/info-cvs
Ineroperability problem with Tortoise CVS
Just spotted a slight interoperability problem with Tortoise CVS and CVS 1.12.9. The new date formatting code confuses the Tortoise cvs log output parsing code as the new local time formatting of dates end up using - to separate the year/month/day fields instead of the / used in older versions of CVS. The TortoiseCVS Version is 1.6.14 which has a CVSNT version of 2.0.41a. Is there anything I can do to persuade CVS 1.12.9 to output dates in the old format? Mike ___ Info-cvs mailing list [EMAIL PROTECTED] http://lists.gnu.org/mailman/listinfo/info-cvs
Re: report on building CVS 1.11.2 on NT
Hi, just to let you know: I've successfully built CVS 1.11.2 on my NT 4.0 sp6a machine using the cygwin tools (which includes gcc 2.95.2). Just following the Unix installation description worked fine. This CVS version also seems to run fine on my NT machine. I've also tried to use VC 6.0, but get stuck there during compilation: nmake /f cvsnt.mak CFG=cvsnt - Win32 Debug snip .\lib\valloc.c(10) : fatal error C1083: Cannot open include file: 'getpagesize.h': No such file or directory Indeed, that file contains lines #ifndef HAVE_GETPAGESIZE # include getpagesize.h #endif Create a windows-NT/getpagesize.h containing: #define getpagesize() 4096 IIRC this will get things to build. Mike ___ Info-cvs mailing list [EMAIL PROTECTED] http://mail.gnu.org/mailman/listinfo/info-cvs
Re: Repairing repository
The device upon which my repository was stored has failed. I have last week's backup, but I had done some checkins since then. Fortunately, I have an up to date tree. I have copied the backup in to create a new repository. Now I must identify the files in the current working tree which are newer than (different from) the files in the repository, and get them comitted. Any suggestions for the easy way to do this? Whatever you do don't do a cvs update. If your local revision is != to the revision in the CVS repository CVS will update it with the repository version potentially losing a lot of changes. My advice for dealing with this is: 1 backup your local working directory. 2 Get a modified CVS which will report an error if the revision number in the repository is less than the one in the working directory (I can supply a source patch for this if you want and possibly a binary). 3 Use the modified client to identify local files that have a revision number greater than the repository revision. 4 For each of these files adjust the revision number in CVS/Entries to match the revision number in the repository 5 Once you have sorted out all the CVS/Entries files you should be able to commit and bring the repository and your workspace back in sync at the cost of a little file history. Mike ___ Info-cvs mailing list [EMAIL PROTECTED] http://mail.gnu.org/mailman/listinfo/info-cvs
Re: Help me
I use WinCVS 1.3.7.1 Beta 7. I want to create a new repository on remote server as the following screen ole0.bmp=20 An error appear in box: cvs init: Empty password used - try 'cvs login' with a real password cvs [init aborted]: d:/StoreCVS/Vnair: no such repository *CVS exited normally with code 1* Okay WinCVS is attempting to access a remote repository using pserver access. I don't think it is possible to create new repositories using pserver. I think yoou will need direct access to the CVS server to create it. Mike PS. Sending binary attachements to this list is a bad idea. If you need to provide this sort of information you would be better off summarising it in the body of the e-mail. ___ Info-cvs mailing list [EMAIL PROTECTED] http://mail.gnu.org/mailman/listinfo/info-cvs
Re: New member and new to CVS
Has CVS ever been used in a client/server environment that includes NT boxes (Windows 2000), and may or may not include a Unix box? Yes. We use a Linux CVS server with primarily Windows clients. There are some unix clients but they are in the minority. There is also an NT based CVS server but I have no experience with it. Mike ___ Info-cvs mailing list [EMAIL PROTECTED] http://mail.gnu.org/mailman/listinfo/info-cvs
Re: using files with .xls and .doc in CVS
1) CVS's handling of binaries is a bit fragile, in that it's fairly easy to accidentally lose the -kb setting. If that happens, subsequent commits will put garbage revisions into the repo (subsequent updates of older revisions, which were committed correctly, will yield garbage working files, but that's recoverable; the garbage commits aren't). I've never actually seen this happen. I have seen people forget to add the -kb flag when adding new files but this can be managed successfully by greating a CVSROOT/cvsrwrappers file which covers all the extensions you will be handling. 2) CVS can't automatically merge changes. This matters more or less depending on (a) how frequently the files change (hence the gif vs. spreadsheet discussion), and (b) how hard it is to merge changes manually, or through semiautomatic tools external to CVS. This can be a bit of a problem. Using watches and cvs edit makes things bearable in this respect. 4) As you noted, deltas for binary files tend to grow more quickly than those for text files, since the former don't diff as well; thus, the ,v files get bigger, faster. Personally, I don't really care; disk is cheap. But others have different circumstances. There is one other thing to consider about this. To do operations on the ,v files CVS reads them into memory. As the file gets larger so does the demand on your CVS server. As long as you don't have really large files that change a lot you should be okay with this for a while. ___ Info-cvs mailing list [EMAIL PROTECTED] http://mail.gnu.org/mailman/listinfo/info-cvs
Re: URGENT: Initial revision number for CVS.
Can it be that the problem lies in the automatic creation of a new branch after new source files have been imported into the repository ? Yes the vendor branch. As soon as you modify the files you will get 2 digit revision numbers. How can I suppress this behaviour ? What people are try to tell you is that you *cannot* suppress this. It is part of how CVS works. If you don't like what import does don't use it; add each file individually and your revisions will start at 1.1. IMO you should learn not to care about revision numbers and start thinking in terms of tags. Mike ___ Info-cvs mailing list [EMAIL PROTECTED] http://mail.gnu.org/mailman/listinfo/info-cvs
Re: Question - commiting modified file
When I generate some files during the build, they appear as 'modified' in CVS. when I do a 'diff' the exit status is '0' indicating that the files are indentical. 1. Do they have 'modified' status due to the file creation time stamp? Yes. 3. After doing a 'commit' on a file that appears modified but is identical, CVS does not check in that file. But the file continues to appear as 'modified'. How can I get it to show the status as current? doing an 'update' will sort things out. You should always do an update before a commit to make sure you are in sync with the repository failing to do so leads to madness and in multiperson projects always some amount of pain. Mike ___ Info-cvs mailing list [EMAIL PROTECTED] http://mail.gnu.org/mailman/listinfo/info-cvs
Re: testing a cvs mirror
Hi I could setup a cvs mirror, but I want to make sure it has all the versions of all the files of the main cvs server. 1. Is there any standard method/tool for testing a mirror? The approach I use is this. Compute MD5 checksums of all ,v files on the master. Compute MD5 checksums of all ,v files on the mirror. Sort the MD5 checksum files and then do a diff. This seems to be a fairly safe approach (CVS is disabled while this check is done so that people cannot modify the ,v files). This will automatically verify all versions of the file at once. This is done automatically just after the mirror has been synced with the main server using rsync. Although rsync has its own checksumming system this has trapped some files that were bit scrambled during the rsync transfer due to bad RAM. The process described takes about 15 minutes on a P200 Linux SCSI system for about 2.5G of data which is a not unreasonable amount of down time in our environment. Mike ___ Info-cvs mailing list [EMAIL PROTECTED] http://mail.gnu.org/mailman/listinfo/info-cvs
Re: some questions about cvs
I am new for CVS. I install a cvs1-11 for Win98 and I use it as command line client to connect to CVS server following the command below: cvs -d :pserver:[EMAIL PROTECTED]:/usr/cvs/dav4j login the error message I got is below: (Logging in to [EMAIL PROTECTED]) CVS.EXE [login aborted]: could not find out home directory How should I figure it out? Thanks You need to set the evironment variable HOME to the name of a directory on your machine in AUTOEXEC.BAT. Adding a line like: set HOME=C:\ will do that. When you log into a CVS pserver CVS needs to create a file to kep the login details in. Unix programs use the home directory to store per user configuration. Mike PS. You will probably get better responses on the mailing list if you send plain text rather than HTML formatted messages. Replying to HTML e-mail is very hard if you have a text only based e-mail client as many people on this list do. ___ Info-cvs mailing list [EMAIL PROTECTED] http://mail.gnu.org/mailman/listinfo/info-cvs
Minor bugfix patch
Attached is a patch which fixes a bug in the Windows command line version of CVS allowing Windows users to do cvs checkout . The patch was developed against 1.10.5 but still applies cleanly to 1.11. Any chance of this patch making it into the main tree? Mike filesubr.patch
Minor bugfix patch
Attached is a patch which fixes a bug in the Windows command line version of CVS allowing Windows users to do cvs checkout . The patch was developed against 1.10.5 but still applies cleanly to 1.11. Any chance of this patch making it into the main tree? Mike filesubr.patch
Re: cvswrappers - any better suggestions ?
Remember that cvswrappers doesn't work if you connect to the repository using a remote protocol (e.g. pserver); you must put a copy of the 'cvswrappers' file named '.cvswrappers' into each users home directory. Actually it does now. It changed somewhere around 1.10. Mike ___ Info-cvs mailing list [EMAIL PROTECTED] http://mail.gnu.org/mailman/listinfo/info-cvs
Re: single-bit errors
This is a multi-part message in MIME format. --=_NextPart_000_0058_01C053F0.8B9CCF20 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable We have noticed that some of our files in CVS are getting off-by-one bit errors. For example, in a text file, an 'L' is converted into a 'K' and a space into a '1' (two bits were modified here). This is not too bad in text files because a differencing tool will usually detect the problem, but binary files could be a more significant problem. Sounds to me like you have a hardware problem with either the client that is commiting these files or more likely your CVS server. The CVS server I maintain had a memory problem recently which on rare occasions caused 1 bit errors in the ,v files. Try swapping out various hardware components until the fault goes away. Our developers are geographically separated, so we use pserver mode where the server is accessible on the internet. Are there any built-in CVS features or extra transmission verification methods we could use to minimize the possibility of errors in CVS? TCP is already quite good at this. However all the network error detection in the world won't help if it is the CVS server that is mangling the files. Mike ___ Info-cvs mailing list [EMAIL PROTECTED] http://mail.gnu.org/mailman/listinfo/info-cvs
Re: Unix to Dos filtering
If I edit a file on Solaris and periodically save the contents, when I open the file on my NT machine using vim I see the most recently saved version (no refresh/remap/logout/reboot necessary. Neither have I. I have never heard of such a thing. Perhaps an application is doing this? Certainly not samba This can happen with certain configuration options (oplocks IIRC) although I thought it was changes on the NT side were not visible to Unix. However configured correctly this should not be a problem. Mike
Re: Enhancement suggestion
I asked if there is an interrest to include my changes to main development branch of CVS under another kflag (e. g. -kkvp). Well even if there's no interest in integrating it back into the main tree I would be interested in a patch which does that. Mike