sun.cc.compiler.txt

1999-08-17T00:00:00
ID PACKETSTORM:11839
Type packetstorm
Reporter Packet Storm
Modified 1999-08-17T00:00:00

Description

                                        
                                            `Date: Tue, 15 Jun 1999 00:17:12 +1000  
From: Darren Reed <avalon@COOMBS.ANU.EDU.AU>  
To: BUGTRAQ@netspace.org  
Subject: big brother in your cc  
  
Whilst this isn't strictly speaking a security bug, it borders on  
Sun acting in a very "big brother" manner which is frightening!  
  
For those of you using Sun's SUNWspro C compiler package, beware!  
The binaries "c89" and "cc" appear to automagically send an email  
to "ut-cc@sunpro.Eng.Sun.COM" with a list of C compiler commands,  
including some sort of cpu-time summary. Extract as follows:  
  
INFO unix i86pc SunOS 5.7  
cc -E  
CPU-time 0.010000 0.010000  
...  
cc -o -Xa -O  
CPU-time 0.000000 0.060000  
...  
cc -o -Xa -O  
CPU-time 0.020000 0.050000  
  
and so on. Setting the environment variable UT_NO_USAGE_TRACKING  
seems to do the right thing but for those that wish to enable this  
feature, check with strings on the available environment settings  
to mediate this (search for UT_). The mail is set to originate from  
"nobody" so it's unlikely you'll notice it if it fails to be delivered  
unless you check your mail queue.  
  
---------------------------------------------------------------------------  
  
Date: Mon, 14 Jun 1999 23:33:09 +0200  
From: Casper Dik <casper@HOLLAND.SUN.COM>  
To: BUGTRAQ@netspace.org  
Subject: Re: big brother in your cc  
  
>Whilst this isn't strictly speaking a security bug, it borders on  
>Sun acting in a very "big brother" manner which is frightening!  
>  
>For those of you using Sun's SUNWspro C compiler package, beware!  
>The binaries "c89" and "cc" appear to automagically send an email  
>to "ut-cc@sunpro.Eng.Sun.COM" with a list of C compiler commands,  
>including some sort of cpu-time summary. Extract as follows:  
>  
>cc -E  
  
(I have a strong sense of deja-vu, wasn't this discussed before on  
BUGTRAQ? Ah wait, Usenet Oct '98)  
  
This compiler "feature" only exists in the pre-FCS compilers (i.e.,  
Alpha and Beta products) and other pre-FCS workshop products.  
  
It was documented in several locations, perhaps even in the  
"must read and agree to" license, but I think it was pretty prominent.  
(The websites have gone now that FCS is here)  
  
(Some older compilers inadvertedly left the code in) 4.0? 4.1?  
  
That is, unless you have a domainname set on your system that ends  
in .sun.com; in that case usage tracking also happens with your FCS  
compiler.  
  
So it's not all that big brotherish as you make it out:  
  
- for alpha/beta only  
- documented how to switch off (in several places)  
- the cc command lines forwarded only include the options,  
not the option parameters or file name arguments.  
(-DFOO becomes -D, -Lpath gives -L etc ; file.c is not listed)  
  
  
Nothing sinister, just alpha/beta users helping to gather statistics  
about compiler option usage. (And us internal Sun folk who get to  
test drive all stuff)  
  
Of course, we could argue whether this should be an opt-in or opt-out  
thing till we're blue in the face, but let's not.  
  
Suffices to say that I've long since disabled most outgoing mail  
>from my system.  
  
Casper  
  
---------------------------------------------------------------------------  
  
Date: Mon, 14 Jun 1999 20:00:05 +0100  
From: Alec Muffett <alecm@COYOTE.UK.SUN.COM>  
To: BUGTRAQ@netspace.org  
Subject: Old Software (Was: Re: big brother in your cc)  
  
[Aleph - please expedite posting this if possible. Love'n'Hugs.]  
  
>Whilst this isn't strictly speaking a security bug, it borders on  
>Sun acting in a very "big brother" manner which is frightening!  
  
Hi Guys,  
  
The story I am told, is:  
  
| This is VERY OLD NEWS. This info was collected as part of the  
| Workshop 5.0 Early Access and Developer Release programs. It is  
| *not*, I repeat, NOT turned on in the FCS release of the product.  
| When customers downloaded the Early Access and Developer Release  
| products off the web they where told this info was being collected via  
| the FAQ and via the web security disclosure statment on the web  
| site. In addition, the FAQ told them how to turn it off if they felt  
| that it was data they did no want to devulge.  
|  
| The Early Access and Developer Release web sites are long since  
| defunct since the WS 5.0 product FCS in 2/2/99. It used to be located  
| at http://access1.sun.com/workshop5.0ea.  
  
...so, can anyone submit an instance of this happening with the non  
early-access software?  
  
- alec  
  
---------------------------------------------------------------------------  
  
Date: Tue, 15 Jun 1999 17:16:52 +1000  
From: Darren Reed <avalon@COOMBS.ANU.EDU.AU>  
To: BUGTRAQ@netspace.org  
Subject: Re: big brother in your cc  
  
I must admit that I'm quite embaressed about bringing this up without  
properly checking which versions, etc, had the described behaviour as  
it doesn't appear in any of the FCS versions.  
  
In some mail from Casper Dik, sie said:  
>  
> (I have a strong sense of deja-vu, wasn't this discussed before on  
> BUGTRAQ? Ah wait, Usenet Oct '98)  
>  
> This compiler "feature" only exists in the pre-FCS compilers (i.e.,  
> Alpha and Beta products) and other pre-FCS workshop products.  
  
Yes, I should have checked more fully on systems I have at my  
disposal.  
  
> It was documented in several locations, perhaps even in the  
> "must read and agree to" license, but I think it was pretty prominent.  
  
And like most licenses which people need to get through to install/get  
software, I (like most people) tend to just click "yes" rather than `waste'  
time reading it.  
  
Still, I'd have rather seen the email come from foo@ rather than  
nobody@ (which has the effect of making it disappear via /dev/null  
if an error occurs in delivery).  
  
Darren  
  
`