- 18 Mar, 2014 1 commit
-
-
Steven Murray authored
-
- 17 Mar, 2014 12 commits
-
-
Steven Murray authored
The CASTOR/SHIFT function sstrerror_r() should call the "standard" function strerror_r(). Before this modification the "standard" strerror() function was being called which is not guranteed to be thread safe. Compiling the CASTOR/SHIFT function sstrerror_r() code as C++ code as opposed to C code makes the GNU compiler check the return type of the "standard" function strerror_r(). The return type of the "standard" function strerror_r() is char * for the GNU version and int for the POSIX version. The following two conditions must be true in order to use the POSIX version of strerror_r(): 1. _GNU_SOURCE must NOT be defined 2. Either _XOPEN_SOURCE should be defined as 600 or _POSIX_C_SOURCE should be defined as 200112L
-
Steven Murray authored
-
Steven Murray authored
-
Steven Murray authored
Clang was clever enough to know that just setting it does not mean that it is used.
-
Steven Murray authored
-
Steven Murray authored
-
Steven Murray authored
-
Sebastien Ponce authored
Do not link with debug, bfd, iberty and z library when not compiling in debug mode. These are useless in that case
-
Steven Murray authored
The sstrerror() function was developed in CASTOR to deal with SHIFT error codes as well as standard errno error codes. The sstrerror() was also supposed to be thread safe. I believe that this is not the case and if I am right then it will be fixed for all of CASTOR in a later commit.
-
Steven Murray authored
-
Steven Murray authored
The CASTOR executables are not running short of CPU cycles. The extra effort required to debug released binaries compiled with the -O2 option is therefore not justified.
-
Steven Murray authored
-
- 14 Mar, 2014 1 commit
-
-
Steven Murray authored
-
- 13 Mar, 2014 16 commits
-
-
Steven Murray authored
-
Steven Murray authored
-
Steven Murray authored
-
Eric Cano authored
-
Eric Cano authored
This was failing in SLC5. The default is now 0/0 insteadof -1/-1 for major/minor in the device list.
-
Eric Cano authored
Added support for TPconfig, looking for drive and density. Improved logging and error situations handling.
-
Eric Cano authored
This is achieved with an extra setup function.
-
Eric Cano authored
-
Eric Cano authored
We now support a new "model" of drives called virtual. This will allow auto-detection of drives in fake environement, for unit tests. Changed accordingly the signature of the reference returned by the Drive wrapper from DriveGeneric to DriveInterface Also fixed the order of the parameters of setDensityCode, to allow the usage of the default densityCode.
-
Eric Cano authored
-
Eric Cano authored
-
David COME authored
-
David COME authored
Added a small section on how-to compile catsor on SLC and possible issues which can be faced while doing so
-
Steven Murray authored
-
David COME authored
-
David COME authored
Exception is now catch by const-ref instead of a copy, message has been improved and move from string::find to string::substr to make sure the file is well formated
-
- 12 Mar, 2014 3 commits
-
-
David COME authored
-
David COME authored
-
Steven Murray authored
The DriveCatalogue::enterDrive() method has been superseded by the DriveCatalogue::populateCatalogue() method. I have therefore removed the now unused DriveCatalogue::enterDrive() method.
-
- 11 Mar, 2014 7 commits
-
-
Steven Murray authored
The contents of the tape drive catalogue can now be populated by passing it the parsed lines of /etc/castor/TPCONFIG.
-
Steven Murray authored
-
Steven Murray authored
-
David COME authored
-
David COME authored
We are now using the matching files and classes frpm castor/tape/threading
-
David COME authored
-
David COME authored
One was requiered because castor::tape::Exception asks for std::string in his constructor
-