Patch testing a ragged issue among 3000s
August 5, 2010
Despite concerns from some independent 3000 software providers, a new patch testing service to replace HP's may not be needed for the community. If there's any demand at all, however, it could surface from an HP action starting in 2011.
Hewlett-Packard said in late '08 it will free up the beta-test patches that it engineered but never general-released (GRed) to the 3000 customer base. Patch testing crawled almost to a stop in the last five years, mostly because HP couldn't get customers to try out most bug fixes and feature enhancements it created while HP's lab still toiled on behalf of 3000 sites.
HP said in a notice to users in the fall of '08:
We will release to the HP ITRC web site the majority of all remaining patches that are still in what we call the “Beta Test Phase." These are patches that have had little or no customer exposure since being developed. [Until 2011] only customers with a valid HP support contract may request beta patches. This strategy... may allow us to test additional beta patches and move them into the General Release Phase.
HP said any beta-test patches which remain untested will be marked plainly when they go into public release. Those extra tests for beta patches since 2008, from support customers? Only a handful of patches got GRed since that announcement. HP's lab closed down two months after the '08 notice.
But ah, there are some custom patches being created for 3000 sites today. Just none we could locate that will alter the system's OS.
One source of patch creation, Allegro Consultants, has been creating patches for Contributed Software Library programs, according to president Steve Cooper. These CSL tools are not usually the kinds of software that interact with independent tools and applications sold by third party vendors, however.
"The CSL is still used and should be hosted," Cooper said. "Changes we make [to the CSL contributions] are things like fixing:
• Check the OS level. If it is 5.0 or earlier, run. Otherwise, abort saying “this is too new a version of MPE, and we haven’t tested here.” We poke it to test for 7.5 or earlier, then test.
• Hardcoded table sizes. It might assume that a PCB entry is 16 words. When it changed to 20 words, the program would hose the PCB and crash the system. So, we would poke in the correct size for newer versions."
Cooper couldn't recall a single recent instance where a customer of his is running a 3000 with a patched OS, where that patch was created by Allegro. "I believe that we have not patched MPE directly in years," he said. "We have patched countless contributed library programs, and other applications, some with missing source code, to work around problems caused by new releases of MPE, by date issues, by larger disk issues, and so on. But I would think that no one is running an MPE system where we’ve patched the OS itself."
There are other support shops available for the smart companies who are planning on independent service for 3000s. But the most likely place to impact the stability of independent 3000 software and applications: HP itself.
A veteran of the community has concerns about the demise of comprehensive testing between patches and these independent vendors, however. "You have little companies that need a patch and have one put on," Paul Edwards said, "and maybe something like Suprtool or Orbit's software stops working. Then those little companies have no recourse." Not because of any lack of tech experience at companies like Robelle or Orbit, but because there's "no liaison between [custom OS patch providers] and the vendors."
Until the HP releases of the beta test patches, though, the issue seems to be a solution in search of a problem. The community has long considered patches a necessary evil; it prefers workarounds for OS problems, and is more likely to have adopted patches that add functionality when it does patch MPE/iX.
The last HP critical patch for MPE/iX came in 2007, when HP modified the OS millicode for the first time in 16 years. HP released a fix for sorted files which involve Large Files and did sorts which in rare circumstances could corrupt data on a 3000. The patch was only needed if a customer's applications accessed mapped files and utilized Large Files -- any which are 4GB or greater in size.