cvs-1.10 -- repository corruption on import?

cvs-1.10 -- repository corruption on import?

Post by Timm Wetze » Fri, 12 Feb 1999 04:00:00



Hello,

has anybody noticed corrupted rcs files in the repository with
cvs-1.10 on NEXTSTEP3.3?
The problem occurs only occasionally (some files in a 50MB kernel
source tree) and does not happen with the same files if repeated.

Scenario: (NS3.3p1, black hardware, source and repository on local
harddisk)
- fresh import of ~50MB netbsd kernel source into empty repository
(- checkout)
- import of newer vendor release (~50MB)
(no error messages here)
- subsequent update/rdiff/... fails because of corrupted file in
repository

The corruptions in the rcs files seem to be mostly in the diff area,
mostly consisting of a block of missing original text and extra null

The cvs-1.10 source is only trivially modifified in xgetcwd.c to use
getwd() insted of getcwd() to build on NEXTSTEP3.3p1.
The binary passed the testsuite (after replacing expr and sh with
gexpr and bash; the last test crashes even my bash).

Is this a known problem with either cvs or NS3.3?
Any code areas to look into more closely?

I'm currently looking into Apple's Openstep/Rhapsody cvs ports but
haven't yet found any interesting differences.

Regards,
Timm
--

Biomed. NMR GmbH        Tel +49 551 201-1091    FAX +49 551 201-1307
Am Fassberg 11, D-37077 Goettingen, Germany

 
 
 

cvs-1.10 -- repository corruption on import?

Post by Charles W. Swige » Fri, 12 Feb 1999 04:00:00


[ ... ]

> The corruptions in the rcs files seem to be mostly in the diff area,
> mostly consisting of a block of missing original text and extra null


That sounds similar to the POSIX append bug.  You didn't compile any part of
CVS, diff, or related tools using -posix or code from the libposix.a library?

-Chuck


      ----------------+-------------------+----------------------------
      You have come to the end of your journey.  Survival is everything.

 
 
 

1. cvs server : cvs co -d ... not working (1.10.5 and 1.10.6)

With the ext method, using rsh, here's what happening :
With a local repository,
cvs co -d . dir
checks out the dir hierarchy in the current repository.

With the same repository accessed by remote shell
cvs co -d . dir
gives
cvs server: existing repository /repository does not match
/repository/dir
cvs server: ignoring module dir

Also, if I specify an absolute path to the -d option, it is ok with
local
repository and gives
cvs [server aborted]: absolute pathname `/users/fimatex/roth/tmp'
illegal
for server
with a remote repository.

The cvs version on the server is 1.10.6, is this bug known and fixed in
1.10.7 ? I also wonder if the client version matters while using rsh :
is it
the client that operates remotely on the repository, or the server that
executes remote orders from the client??

TIA

Michel-Ange CAMHI
Ingnieur systme - ATOS Multimdia
ATOS - ZI La Pointe - 59113 Seclin
Tel: 33+ (0)3 20 60 7896 - Fax: 33+ (0)3 20 60 7649 -

2. named: connection refused

3. CVS: repository corruption

4. Slide Feeder SF-200S for Coolscan 4000 ED available!

5. new CVS repository without import

6. Print delay from shared printers

7. Q: How to import a complete project from one CVS repository to another?

8. Selling Palm V Modem

9. importing newer CVS sources into old repository

10. CVS 1.10.5 and 1.10.6

11. patch for test cvs-1.10 needed

12. CVS: What does 'P' mean on update - CVS 1.10.3 WinNT Client

13. can't build CVS-1.10 on Solaris 7 x86