Re: Thread safety clarification

[David Waite <mass at akuma dot org>]

Loren James Rittle wrote:

>Benjamin covered the rest of your e-mail (in terms of asking for more
>information ;-).
>
yep - always helpful to know where to start :-)

>>According to the FAQ in CVS, you cannot assume that two separate threads 
>>can simultaneously read from a STL container class at once; however, 
>>according to the containers howto (23_containers/howto.html) and the SGI 
>>multithreading document (http://www.sgi.com/tech/stl/thread_safety.html) 
>>both say this is acceptable behavior.  Which goes - is the lack of 
>>thread-safety on container read an issue with libstdc++, and if so, what 
>>is the actual issue (as perhaps I can work on it)
>>
>
>For STL components, IMHO 23_containers/howto.html and SGI documents
>(excluding any configuration information given there) override
>anything you may have found elsewhere.
>
>In particular, the concrete example in faq/index.html (5.6 Is
>libstdc++-v3 thread-safe?) explicitly involves mutating a shared
>object and is not really the same thing you ask about.  Since we are
>talking about the entire library, we must be conservative in the
>explanation of the design rules.  However, I do agree that the
>follow-up text needs some additional clarification to avoid being
>overly conservative:
>
>      [...] This requirement includes both read and write access to
>         objects; do not assume that two threads may read a shared
>         standard container at the same time.
>
>How about the following text?
>
>     [...] In general, this requirement includes both read and write
>         access to objects; unless otherwise documented as safe, do
>         not assume that two threads may access a shared standard
>         library object at the same time.
>
>This change works by generalizing the word choice to ``standard
>library object'' from ``standard container'' while making it clear
>that there are exceptions.  When one does more detailed study in the
>23_containers/howto.html, one will see that readonly access is
>well-defined on the standard containers in some conditions (i.e. a
>documented exception to the general rule).
>
Much better :-)

Are there examples of objects which cannot be accessed from multiple 
threads simultaneously? For example, can I have multiple threads reading 
from a globally-defined, constant string?

-David Waite