From: Russel W. <ru...@ru...> - 2001-12-14 10:06:34
|
Yesterday 2001.12.13 (well actually the day before 2001.12.12 but we didn't follow it up till yesterday) we found that our C code which had previously compiled and run correctly with SDCC compiled but did not execute with the same behaviour -- indeed failed to run at all really. Given that this code of ours has remained unchanged over the last week, we initially assume a problem with SDCC. It is true that the possibility is that our code has always been incorrect and that it was a fluke that it worked previously but we are treating this as the backup hypothesis for the moment. We pulled a version of the code from "2 days ago" and this exhibited the same problem. We pulled a version of the code from "4 days ago" and the 'problem' had gone away. Assuming a sensible dating system, we are therefore looking for a problem introduced between 2001.12.09 and 2001.12.11. Clearly this is insufficient information to track a bug, we will do more work on differences in the CVS tree -- pulling copies of the tree by date/time in order to locate the update that causes the problem. If however, anyone can remember the semantics of the changes made in those two days it could help our syntactic tracing. This leads to another issue. I sent in to the list a set of rather trivial, but crucial to us, changes to as/mcs51/lkmain.c and src/SDCCmain.c but the changes have not been actioned or even commented on. This means we are currently working with modifications in our copy of the tree which is not good. Unless I have missed something, there appears to be no formal process for people without write permission to the master tree to submit changes, have them validated and then entered into the CVS tree. My worry is that if we spend time tracing the current problem in compiling/executing our code, we will end up with a set of fixes that will not be entered into the master CVS tree. If this is the case then we would probably end up having to declare a branch so as to get the safety of CVS history but I would much rather not do this since branches should be used for technical reasons not management reasons. I guess what I am asking for is an update process that is open and works. -- Russel. ===================================================================== Dr Russel Winder +44 20 7585 2200 41 Buckmaster Road +44 7770 465 077 London SW11 1EN, UK ru...@ru... |