Status for Andrew DeFaria

• Portgd still not working well
• Build 3.0.0 TOT again
• Added package of pdn to 2.1.0. Somehow I missed it
• Repackaged cdklinux

Portgd's continuing woes

Turns out the problem is that we are attempting to do a cvs add from outside of the repository. Portgd works from one directory and checks out an los178.from and an los178.to. Then from that directory it attempts to cvs add los178.to/<path>/<dir>. This will not work. We need, instead, to cd los178.to then do cvs add <path>/<dir>.

But portgd was also having a problem trying to recursively add the directories above new files. How to debug/test this is the question...

Posted by at 5:00 PM | Permalink

• Changed portgd script to attempt to handle new CVS files/li>
• Changed package.sh for 3.0.0 to package NetSNMP
• Release LOS178 2.1.0. Still working on building GDB on Windows
• Built and released LOS178 3.0.0 20050825. Turns out the toolchain I was using does indeed have the latest!
• Port of GD from LOS178 has some missing CRs. Will work on this tomorrow

Posted by at 12:43 PM | Permalink

• Performed a TOT 3.0.0 build. More stuff coming in...
• Struggling with 2.0 GD build. Turns out I was missing some CRs. Finally building!
• GDB Server build on Solaris is OK now but it fails on Linux. I suspect a packaging problem
• The portgd script has issues

Portgd Challenges

One of the problems with portgd is that the from CVS repository really needs to be set to HEAD because you want the latest but you're gonna pull the CRs by number and individual file revisions. This may cause problems sometimes (think ordering of CR numbers where pulling one CR overwrites a previous one in the wrong way. BTW This is why branching is good!)

Additionally portgd has a problem where the CR introduces new elements (Clearcase term - think new files and/or directories). When portgd attempts to cp the file to the to CVS repository it may fail because the parent directory did not previously exist and is not in the to CVS repository.

The portgd script can be updated to check to insure that each directory in the path of the file revision to be "ported" exists and if not then to create it and add it to CVS. It should also check to see if the file is new and again copy it but also add it to CVS.

Posted by at 11:32 AM | Permalink

• Linux ABI build fails. Reported to vlad
• GDB build failed. Rebuilding...

Linux ABI build failure

The Linux ABI fails to build on yellowdog Linux.

    + DebugFlags=
    + export GLIBC_LOCATION LYNX_PREFIX LIBLYNXPTHREAD_PREFIX
    + export LYNXOS_VERSION_MAJOR LYNXOS178_VERSION_MAJOR
    + '[' 1 '!=' 0 ']'
    + '[' '!' -r /home/int/LOS178-300/20050822/linuxabi/glibc-2.3.3/../liblynxpthrea
    d/Makefile ']'
    + echo ======================= BUILD - LIBLYNXPTHREAD ==========================
    ===========
    ======================= BUILD - LIBLYNXPTHREAD =================================
    ====
    + cd /home/int/LOS178-300/20050822/linuxabi/glibc-2.3.3/../liblynxpthread
    + make PLATFORM=ppc
    make: *** No rule to make target `include/itimer.h', needed by `all'.  Stop.
    + exit 1

itimer.h has been removed. There are other problems too.

GDB Build Failure

For some reason it turns out that several libraries such as libc.a appear in the dev area as if they were built as production. IOW the size of libc.a(dev) == libc.a(pdn). Don't know why this happened because the install.log for dev seems to indicate that it was built in dev mode. May be a problem in the build_los178 script. Rebuilding...

Also, due to hardcoding of bison's location we must build in /usr/los178/2.1.0/ppc_dev.

Posted by at 6:10 PM | Permalink

• Continuing with LOS178 2.1.0 build... Built CDK on Solaris, Linux and Windows
• Worked out a bug with files4[e]cr where it has problems with newly created files

Posted by at 3:13 PM | Permalink

• Built and release LOS178 3.0.0 20050818
• Built and release HybridOS 2.0 20050818
• Need to build LOS178 2.1.0. There's also a a GDB 6.0 build coming up
• Need to learn how to build KDI. This will become the last step - a sanity check for the build

Posted by at 6:48 PM | Permalink

• Built Ada compiler with a 20050623 devos base
• Built and released LOS178 3.0.0 20050817
• Changed build_los178_343 to have a new option that allows the build to continue if the CDK build fails. Sometimes we still allow some CDK build failures

Ada Build

I managed to get Ada built (with a few procedural errors that Jesse's looking into) on hornet using the 20050623 devos. Note that the process requires a 3.4.3 or greater native toolchain. Luckily there is one available on t3, however the following symlinks were required:

    $ mkdir -p /usr/products/compilers
    $ cd /usr/products/compilers
    $ ln -s /net/t3/export/tools_prod/products/compilers/FSF FSF
    $ ln -f /net/t3/export/tools_prod/products/compilers/FSF/gcc-3.4.3 latest

Also note the usage and reliance on amd and the -hosts map. Finally the build process expects the "latest" link but the compiler itself expects FSF.

Jesse is also looking into what would be required to build utilizing a later devos.

Posted by at 6:59 PM | Permalink

• Changed error handling in build_los178_343 to be more resilient
• Worked with Jesse on Ada build problems
• Added some documentation to mirage/docs

Posted by at 6:54 PM | Permalink

• Ported changes from LOS178 20050801/20050806 -> GD
• Built TOT for LOS178 3.0.0

portgd

Was asked to port the changed from LOS178 -> GD again. As this is a frequent occurance it screams for automation. As I had already coded a verifycrs, which essentially automates verification of ported [E}CRs from one repository to another, it seems a simple case to derive as script to do the porting. Coded up portgd. Note that portgd does not perform automatic building - just porting of the files associated with the passed in [E}CRs.

build_los178_343

Changed build_los178_343 to go back to using scp to obtain the toolchain tarball. This is still kludgy because it entails hardcoding the location (which is date based) of the source tarball. However, using NFS requires that the NFS area is accessible and in the past it was an automount point to an assumed -hosts map location. That doesn't work on Linux systems using autofs (an investigation of using amd instead of autofs has not yet been performed). As such, from a Linux system, the toolchain tarball was not easily accessible via NFS. So we're back to using scp, which has it's own problems, namely the assumption that passwordless scp access has been set up.

Posted by at 11:50 AM | Permalink

• Rebuilt LOS178 2.1.0 with tar file from Sergey
• Moved a few more functions into BuildProcs.pm so that both build_los178 and build_los178_343 can share them
• Filled out build_los178 to do packaging, tagging and the like that build_los178_343 does

Posted by at 5:13 PM | Permalink

• Built 2.1.0 TOB. Production still fails but dev builds so we packaged that

Posted by at 8:09 AM | Permalink

• After setting up much of Mirage we had problems with the disk. Luckily Vinnie was able to recover it.
• Setup Ecrdig on Mirage. For some reason Mirage lacks Net/Sockets.php. Copied from Saturn. It works!
• Setup CVS Reports on Mirage
• Added some content and arranged things a bit

Posted by at 8:16 AM | Permalink

• Started installing things like cvsadm, cvsreport and phpfaq on Mirage when poof! the /home filesystem got corrupted! Looks like we might need to start over again!

Posted by at 3:25 PM | Permalink

• Cleaned up some more on build_los178_343. Basically expanding Logger.pm such that it logs the detailed output to the logfile but echos out just the execution of the command line to stdout if verbose. This allows the logfile itself to contain more detailed output such as the output from make install and cvs tag's etc but the screen isn't overrun if one does a build_los178_343 -v.New webserver finally here! Moving stuff over there and configuring things

Posted by at 8:23 PM | Permalink

• Moved some routines from build_los178[_343] -> BuildProcs.pm.
• Need to build LOS178 with Toolchain 3.4.3 first or else build of Ada doesn't even try

Posted by at 12:55 PM | Permalink

• Moved mail_log -> Logger.pm as a method named mail. Now you can easily have a log object mail itself!
• Added Logger::logcmd which runs a command logging output into the log file
• Changed build_los178_343 to checkout toolbox and tst since we need them for packaging anyway
• Got packaging and renaming to work
• Changed chdir's to use $base to make getting around easier
• Worked with Adam and Jesse on building the Ada compiler in HybridOS

Posted by at 6:11 PM | Permalink

• Added tagging to build_los178_343
• Started looking into building gnat on HybridOS

Posted by at 4:52 PM | Permalink

• Started converting build_los178 -> build_los178_343 as the 3.4.3 toolchain based build works differently

Posted by at 4:59 PM | Permalink

• Following Vinnie's instructions, built and released LOS178 3.0.0 20050801

Posted by at 6:29 PM | Permalink