[ts-gen] compile problems on cygwin

R P Herrold herrold at owlriver.com
Sat May 17 10:11:42 EDT 2008


On Fri, 16 May 2008, Gabriel Ricardo wrote:

> Bill and Russ, Thanks for the quick response. For some 
> reason I'm not getting emails from the list, I have to go to 
> the archive to read your responses (perhaps this is because 
> I just joined?).

curious -- I saw your subscription notification come through 
yesterday; it _should_ be offered to Yahoo's mailservers

> It appears you guys found the culprit pretty quickly.  It 
> would have taken me forever to get to the bottom of this. 
> Unless you guys know of a workaround, my next step is to try 
> using gwc (gcc for windows).

Bill is the one with the hypothesis, and the ability to write 
a nice minimal test; me, I'm usually just googling the error 
mesages, and poking fixes into code.

The other workaround it so run a virtual machine 'nix instance 
inside a Windows box, but that is probably not what you seek. 
The shim, and of course the TWS and MySQL are all 'socket 
based' applications, and so one can run them on a mix of 
platforms trivially.

The last three weeks, on the internal developmental releases 
since the 4/11 external, I have been running against a OS/X 
TWS, and with a database server on a 32 bit Linux box 
different than the usual 64 bit box I formerly ran the 
database on, and a 32 bit shim instance (there is a problem I 
exposed doing the zLinux port in the ./lib/ code word size, 
and we temporarily have blocked 64 bit as there are 
potentially erroneous states the shim can work itself into 
with large enough datasets.

I think the shim users community is going to ** really ** like 
our next external series.  We are close to having Nils' order 
modify (and cross session restart order management) issue 
really and truly solved with a redesign, and have added some 
other special treats to get us full symbol set coverage in all 
markets.

More important from a robustness standpoint, we largely (I 
would say 'completely', based on the design work, but I have 
not seen the final 'as implemented' product to _know_ that we 
made it) removed the need for the end user to directly write 
into the database when the shim is running, by having the shim 
assume some added behind the scenes functions.  This greatly 
simplifies shim usage by the downstream client(s).  Many 
synchronization issues go away.

We do not stress it, but one can presently have one or more 
--data mode clients (up to 7), and one --risk mode client to 
the shim simultaneously [permitting tick and history 
harvesting], right along an actively trading instance.

> Russ, I believe you are correct, bugs in cygwin/gcc are 
> reported to cygwin at cygwin.com

Yup -- I found a release notice from that project, poked 
around a bit more, and filed a report already.  Thre seems to 
be an unwritten rule there not to post the 'harvestable' email 
addresses of others into their mailing list which I (sadly) 
innocently violated.  ;(

I'll try to shave down our code to get to the minimum 
reproducer for them next week.   As you note, the code is fine 
under Linux, and also under OS/X, so I have to think there is 
some dark corner in the Windows implementation which we are 
tripping into.

-- Russ herrold


More information about the ts-general mailing list