If you don't agree that the dependence on Cygwin is hampering eCos,
then don't read further.
Thus warned.....
Problems w/Cygwin:
- Whether or not Cygwin will work at all for eCos is a matter of
chance depending on the mix of Cygwin packages on any given day. E.g.
today they seem to be battling with breakages in heap allocation on
XP due to trying to accommodate Vista somehow. Who knows what they'll
be working on tomorrow. Cygwin ecosconfig binary available for
download on ecos.sourceware.org was broken for years(?) w/latest
Cygwin. http://www.cygwin.com/ml/cygwin/2006-01/msg00799.html
- There is no way to create a single package no questions installer
w/Cygwin. Lots of companies ignore Cygwin license conditions and
bundle cygwin1.dll which creates lots of impossible to debug problems
w/side-by-side installs. A pox on them!
- Installing or upgrading Cygwin is *painful*. Download times,
unpacking times, lots of questions, how to figure out which packages
to install, etc., etc.
- Cygwin has no "release" concept a la Linux distributions. I don't
think there is a single change (except bugfixes) that has interested
me in Cygwin the last 5 years w.r.t. eCos development. eCos really has
very modest needs. I believe the reason this is so, is because there
is no interest w/the Cygwin package maintainers. The only option I
know of is the Cygwin Time Machine. For the record: I don't believe
the Cygwin guys are doing anything wrong. Doing releases would be the
solution, but to do that the resources would have to come from
somewhere. Given their resource limitations, I believe they are
spending their resources wisely.
As far as I know, MinGW binaries of GCC toolchains + eCos tools is the
most tantalizing prospect.
Since a eCos needs are so simple & unchanging, I believe a MinGW based
package could live relatively unchanged for years.
Does anyone know of any organized efforts to address these problems?