[Scons-users] Scons-users Digest, Vol 5, Issue 21

[John Galbraith]

Bill, this is a good suggestion and one I had not thought of.  It is not 
quite as elegant a solution to my eye, but it still works fine and is 
robust to crashes.

Also, it occured to me that true crashes are actually less frequent than 
my own kill signals, sent because I want to change something before the 
run finishes.  In this case, I might be able to catch the interrupt 
signal and shutdown gracefully.  I have done this exact thing in C++, I 
just have to figure it out for python.  I think the hard part is that 
you do not know in advance which thread gets the signal.  Signal 
handling and threading do not necessarily play nice, at least on Linux 
which is the only platform I know.

I still could totally go for a env.flush() mechanism...

John

On 10/28/2012 02:52 AM, scons-users-request at scons.org wrote:

> On Oct 24, 2012, at 10:09 AM, John Galbraith<jgalb at lanl.gov>  wrote:

>

>> >

>> >Is there any way for scons to protect itself from broken extension calls, and at least flush .sconsign before surrendering to no-fault flaming death?

> So you're pulling your c extension into your SCons invocation and running it in process in SCons?

> Sounds like generally a bad idea.

> Make it run in a command run by SCons, that way if it crashes, SCons will still write a proper .sconsign

>

> -Bill