Freddie Chopin wrote: >> But I disagree with reverting the commit. Fix the problem instead. > > What? > > So you don't know how to fix that, I don't know how to fix that, but > we cannot revert a one-line change that introduced A BUG just > because it (supposedly - I did not check) adds some functionality > that was not actually discussed in detail EVER?
It is obvious that the commit is a significant improvement for life in Tcl files, that needs no discussion. Thus, we really want it. We can also assume that Steve did not pick this way to do the improvement because he likes to bully Windows users, but because it is the better way to reach the goal. We have discovered a side effect of uplevel, so we should find out how we can deal with that side effect, not abandon the improvement. > Why can't we revert a change, fixing a bug and wait for a real > jimtcl-expert (Steve) to discuss other alternatives Maybe you, like me, roughly know the capabilities of Tcl, even if you are not an expert? It is a pretty basic language, so it is fairly safe to assume that there is no alternative to uplevel. > while in the meantime OpenOCD would work as expected for Windows > users? Of course I agree that it would be great to fix the problem quickly. But abandoning the improvement is not the answer. > "The problem" is that Windows paths now cannot be used normally and > OpenOCD cannot be used as before. It is trivial to work around this problem. Open src/startup.tcl and remove uplevel #0 from line 58, and you are done. Since src/startup.tcl is generated during build, git will not care that it has been changed. Of course the workaround needs to be re-applied in case one of the .tcl files that make up startup.tcl changes, by your hand or by a commit you fetch. > Global variables thing which this patch (supposedly) improves is > just an improvement, because you could use globals before anyway, > just had to add "global NAME" before. I think it's a very desirable improvement, and even though it does not directly affect what you and I have so far focused on in OpenOCD (ie. not .tcl files) I guess you agree that it can make a difference for those who do work with .tcl files. More importantly, every user is exposed to the .tcl files, so anything that allows to simplify or clarify them is real important. > Please revert to fix a real problem (removing backslashes from > Windows paths). This is not the problem, this is just how double evaluation manifests itself. Since when is treating the symptom any good? > We can deal with improvement while at the same time OpenOCD works > as it should. Please do deal with the improvement right away. I don't expect it will have to be very complicated. Meanwhile, the workaround is quite trivial. I'm more interested in knowing if not using uplevel also changes which directory relative paths are relative to? //Peter _______________________________________________ Openocd-development mailing list Openocd-development@lists.berlios.de https://lists.berlios.de/mailman/listinfo/openocd-development
