« January 2005 | Main | March 2005 »
Taking code from cqd/cqc that I implemented over at Salira, I decided to adapt these to ECRs. The concept is essentially the same - you have data on a server in a database and wish to get to it either through a command line or a web page. At Salira I had Clearquest and it's backend database of SQL Anywhere. I was also limited to using cqperl which was based off of ActiveState Perl which has problems going into daemon mode and handling signals. Here I have an Informix SQL database and do not face the same limitation of ActiveState. Perl's DBI interface makes handling this database pretty easy. Of course the data fields are different and I took advantage of real Linux so as to make the daemon mode actually work well now. Multithreading remains a problem in that the child process hangs on the SQL prepare statement and if I try to reopen the database (not efficient) I get an error. Eventually I can work out these problems...
The solution was simple but it was also not ideal. The expect scripts currently merely fill a few arrays with numbers from ranges like 18000-18999, 19000-19999, etc. Then, later on they just grab a number from the array.
The idea of adding "-xx" to the end of the product numbers is to allow for different versions down the road. So one might have 18000-00.devos.tar.gz and 18001-00.other.package.tar.gz. Other times there might be 18000-00.devos.tar.gz and 18001-01.other.package.tar.gz. IOW different "-xx"'s in the product range. Unfortunately the way the expect scripts are coded this is not easy to do.
So for now we are just implementing this by adding -00 to the end. IOW all -xx's must be the same, for now...
ECR 23248: Toolchain build fails due to problem in LynxOS header file:
In attempting to build top of the gcc_3_2_2-branch gdb fails to build with the following:
Spent most of the day attempting to build the TOB Toolchain by first tagging it with DEV_LYNXOS_3P2P2_ALL_20050207. The tagging operation has some problems and I suspect not everything that needed to get tagged got tagged.
Have been hunting down and tagging things hopefully appropriately in order to get things to build but I'm still having some problems.
Changed cvs_report.pl to:
Changed http://saturn.lynx.com/cvsr to generate HTML from the above datafiles.
The changes to cvs_report.pl involved changing ProcessECRs to look for the start and stop revision numbers in the log. The CVS log has entries like the following:
$ cvs log Makefile | grep ^revision revision 1.9 revision 1.8 revision 1.7 revision 1.6 revision 1.5 revision 1.4 revision 1.3 revision 1.2 revision 1.1 revision 1.1.2.4 revision 1.1.2.3 revision 1.1.2.2 revision 1.1.2.1 revision 1.3.4.1 revision 1.3.2.2 revision 1.3.2.1
Thus if the from tag is pointing to revision 1.1.2.1 and the to_tag is pointing to revision 1.1.2.4 cvs_report.pl now skips all versions until 1.1.2.4 then picks up ECR numbers from there.
First check out the top of trunk of ats:
steps_com=" 1 2 3"; echo "=== Starting common ===" startit.exp common dummy $steps_com >> LOGS/common.log 2>&1 &
Either I really just don't understand CVS branching or perhaps we have an error in tagging. I have written a Perl script that will report the ECRs involved between two different CVS tags. When I run this script on the toolchain, comparing tags Lion_com3_2_2_120204 -> Lion_com3_2_2_121604 I see odd results. I believe that the only real ECR involved here is ECR #23084. The following appears in Quintus for that ECR:
/cvs/gcc-cvs/toolchain/src/ChangeLog.lynx,v <-- ChangeLog.lynx new revision: 1.25.2.31; previous revision: 1.25.2.30 /cvs/gcc-cvs/toolchain/src/gcc/config/lynx.c,v <-- lynx.c new revision: 1.2.2.1; previous revision: 1.2 [Check-ins] on trunk /cvs/gcc-cvs/toolchain/src/gcc/ChangeLog.lynx,v <-- ChangeLog.lynx new revision: 1.7; previous revision: 1.6 /cvs/gcc-cvs/toolchain/src/gcc/config/lynx.c,v <-- lynx.c new revision: 1.3; previous revision: 1.2
However I also notice the following WRT the file src/ChangeLog.lynx:
$ cvs log toolchain/src/ChangeLog.lynx
...
Lion_com3_2_2_121604: 1.25.2.31
...
Lion_com3_2_2_120604: 1.25.3.28
...
revision 1.25.2.31
date: 2004/12/15 23:48:49; author: anemet; state: Exp; lines: +6 -0
ECR 23084
...
revision 1.25.2.30
date: 2004/12/14 15:28:22; author: orepin; state: Exp; lines: +17 -0
ECR 22875
...
revision 1.25.2.29
date: 2004/12/14 09:26:25; author: orepin; state: Exp; lines: +8 -0
ECR 22805
...
revision 1.25.2.28
date: 2004/11/23 15:35:16; author: aneyman; state: Exp; lines: +8 -0
ECR 22989
...
$
So, to me, the ECRs that changed between the CVS tags of Lion_com3_2_2_120204 and Lion_com3_2_2_121604 WRT src/ChangeLog.lynx would be
(I assume that ECR #22989 was included with the previous toolchain).
Or am I just looking at this the wrong way?
Futhermore, the other files associated with, for example, ECR #22875, are:
Yet, other than ChangeLog.lynx, all other files have tags that suggest that nothing else changed! For example, gdbtypes.h has Lion_com3_2_2_120204 pointing to revision 1.2 and Lion_com3_2_2_121604 also pointing to 1.2! Likewise for gdbtypes.c and config/i386/tm-i386lynx.h. Is this a tagging error?