Menu
▾
▴
#2 Test tool packaging and distribution
I'm wondering how STAF users go about distributing and
installing a test tool (executables and associated
datafiles) to several systems before they use
the 'staf systema process start ...' to start the test.
It seems that there are several ways:
* Have a script that runs 'STAF ... FS ...' to copy
the files over; followed by a 'staf ... process start'
to install the files of the tool.
* Perform tool installation without using STAF.
I'm just wondering if there are other solutions or
techniques to solve the problem.
Logged In: YES
user_id=279901
That is a good question. Most groups simply keep the tool
in a zip/tgz file. Then they just transfer the package
(using STAF FS service) and unzip/untar it on the other end
(by starting a process via Process service). Some groups
have created more sophisticated mechanisms, but have not
made them available to the community as a whole. We still
have an old Rexx service running around which we called
FileRep. It wasn't much more than a simple wrapper around
the FS/Process combo mentioned above. We haven't open-
sourced it because it hasn't been used in a while and we
would need to do some significant updates to make it ready
for the open-source community.
In the long run, I think some kind of Package Manager
service would be great. Now, if we could just find someone
with spare time on their hands. :)
I'm going to keep this request open in case other folks
have anything to add, or until a better answer (such as the
service) comes around.
Logged In: YES
user_id=37866
Is RPM portable enough -- it is a very good packaging
system, and I would be willing to write a STAF RPM service,
esp since it would help us out as well...
Advantages: Well-known, well-understood, widely used, easy
to embed
Disadvantages: Introduces an external dependency, may not be
ported to all targets
Logged In: YES
user_id=63714
RPM would be a nice solution - as long as it runs on
MS-Windows too. (I haven't heard of it being ported to
MS-Windows, but I haven't been monitoring RPM lately).
Logged In: YES
user_id=279901
As jbd mentioned, RPM would probably be good enough so long
as it ran on a sufficiently wide variety of systems. I
would think Windows, Linux, AIX, and Solaris (and maybe HP-
UX) would be sufficient. The real trick is that the type
of service described here is one that would preferably run
on a single system, yet still be able to distribute to a
variety of systems. So, for example, I could run the
service on a Win32 box and have it distribute out testcases
to Win32, Linux, and Solaris boxes (without those systems
needing to have the service installed).