"Scalability" Issues with the libgcj Makefile

Jeff Sturm jsturm@one-point.com
Tue Aug 27 20:41:00 GMT 2002


On Tue, 27 Aug 2002, Ranjit Mathew wrote:
>      As an update on the "libtool" and "ar" issue, I
> must mention that it was "solved" when I changed
> the value of "max_cmd_len" throughout the libtool
> script to something like "3000" - it then used
> "piece wise linking" to create the final libgcj.a.

Then I wonder why ltconfig would've computed a max_cmd_len longer than the
MSYS shell can handle?  Or perhaps it's a limitation of ar?  I really have
no idea how command lines are handled by MSYS.

> Another "scalability" issue (stack dump with the MSYS
> make) came when I tried to install the final
> GCJ/libgcj combo. Specifically, the problem was
> in the "install-data-local" target of the Makefile
> and was once again due to the use of $(shell ) within
> a $(foreach ) that had a lot of header file names to
> process.

This needs to be fixed in make.  As ugly as it is, that $(foreach) is
necessary to avoid long shell commands.

Jeff



More information about the Java mailing list