Release Notes and Readme useless for 10.1.014 (Win)

[levicki]

I just downloaded 10.1.014 and first thing I did was to check the readme file w_cc_p_10.1.014_README.txt which says:

***********************************************************************
This file will be updated soon with descriptions of problems fixed in this release.
***********************************************************************

Date this file was posted is February, 1st. How soon is soon?

Then I read the w_cc_p_10.1.014_Release_Notes.htm which says:

Change History

10.1.012

• Corrections to reported problems only

10.1.011

• /arch:ia32 added

• /Qopt-jump-tables added

• /Qopt-loadpair[-] added

• /Qopt-mod-versioning[-] added

• /Qopt-prefetch-initial-values[-] added

• /Qopt-prefetch-next-iteration[-] added

• /Qunroll-aggressive[-] now available on IA-64 architecture

10.1.008

• First release of 10.1

Even the 10.1.013 changes haven't been documented in there not to mention that the string 10.1.014 can only be found in the file name.

What kind of useless documentation is this?!?

• Release notes which do not mention two latest releases?!?
• Readme which has nothing to read?!?

Seriously, what kind of crap is that?!? Did anyone proof-read what has been posted for download?!? It is there for 8 days for Christ sake and nobody is bothered enough to fix it?!?

What should I tell the company I work for if they ask if the issue I submitted via Premier support has been fixed in the latest compiler build?

They won't authorize me to waste any more time uninstalling the old compiler, installing and configuring new one, and testing for the same compiler errors over and over again just to see if they were fixed by chance!

When you are selling a product this is not exactly the right way to support it.

[Dale_S_Intel]

Sorry for the delay, Igor. Apparently there was some hangup with the issue list, but they decided it would be better not to hold up the release just for that. The readme.txt file has now been updated.

As for the "Change History" in the "Release Notes" file, I suspect that refers to changes in that document, i.e. there were no changes in the 10.1 Release Notes in 10.1.013 and 10.1.014, so they're not mentioned there.

In general, though, you should be notified by a note on your issues in premier.intel.com (typically also by email when a note is added), so ideally you shouldn't have to check the readme.txt file for every release to find if a particular issue has been changed. If that's not working for you, please let us know.

Thanks!
Dale

[levicki]

Dale,

Change history doesn't refer to the document changes, but to the compiler changes. If you actually read the excerpt of the change history I posted above, that should've been obvious to you. One doesn't add compiler switches to the document but to the compiler.

As for the premier.intel.com, support people always say they will notify me, but I often discover the changes myself much faster (sometimes a month or two in advance of their notice).

Sorry for being a bit harsh here, but I believe this should never happened -- I really didn't like having to report it.

[Steven_L_Intel1]

I apologize for the problem with the release notes. This appears to be my fault though I don't understand what happened. I'll investigate further. The updated release notes are attached to this post.

As for the README, an updated version is now available from the Intel Registration Center.

[levicki]

Steve,

Thanks for the updated release notes, but the other customers getting it from Registration & Downloads Center still have a stale copy here.

I am not only surprised but also severely disappointed to find out that the issue #458830 (overloading new operator does not allow aligned allocations) which I submitted on 11/15/2007 still hasn't been fixed after 60 days.

Based on my personal experience, Intel compiler developers manage to fix about 4 bugs per year out of those I report. With that speed we aren't getting anywhere.

[Steven_L_Intel1]

Igor,

I have requested that the correct release notes be provided at the Intel Registration Center.

I looked up your issue and saw a comment from the developer saying that making this change would break link-compatibility for other objects and that the change was deemed too risky to put into a maintenance update that gets no external testing. It has been fixed for a future release that will undergo extensive beta testing. Feel free to update your issue at Intel Premier Support if you disagree with this decision. I do note that the support engineer did not communicate that to you and I will bring that up with him.

As of version 10, we have put into place a process where not all bugs automatically get fixed in maintenance updates as the risk of regressions with customer applications is higher. We take into consideration the seriousness of the problem, whether there is a reasonable workaround and the risk of making the change when deciding which fixes to include in an update. Decisions are open to negotiation and reconsideration - your input is an important part of that.

[levicki]

I looked up your issue and saw a comment from the developer saying that making this change would break link-compatibility for other objects and that the change was deemed too risky to put into a maintenance update that gets no external testing.

Steve, I would really like to know — how is it even possible for such a grave coding error to make it into a tested release?

Feel free to update your issue at Intel Premier Support if you disagree with this decision

I cannot disagree with something I haven't read yet. Furthermore, since I am not able to test it, I don't know if it would break anything or not. In my opinion, they should also explain how that fix would break things.

Finally, if this kind of error can get into the official release, then what is the point of not allowing the fix to get into the maintainance release?

[Steven_L_Intel1]

Igor, I suggest that you update your Intel Premier Support issue with your comments. I'm a "Fortran guy" and don't understand the details of the issue as well as a C++ expert would.