From: Gilles E. <g....@fr...> - 2006-07-31 09:48:22
|
Selon Franck Bourdonnec <fbo...@ch...>: > Le lundi 31 juillet 2006 07:13, Gilles Espinasse a =E9crit=A0: > > ----- Original Message ----- > > From: "Ivan Kabaivanov" <ch...@ya...> > > To: "Franck Bourdonnec" <fra...@us...> > > Cc: <ipc...@li...> > > Sent: Monday, July 31, 2006 3:52 AM > > Subject: [IPCop-devel] tftp and www in make.sh > > ... > > > > Sure, I have my own script to do that but I don't see a gain to force > > everyone to fill that part if he don't care about pxe, even it is ver= y > > practical during testing. > > > > Gilles > > > FORCE? WHAT FORCE?? it is an option. > there is nothing to answer if you have nothing to answer here. Change > the text as you want if unclear (bad english) but let make.sh > do a complete cycle of compile/build/copy to the write place. > There was more or less 5 questions that you was forced to answer every ti= me ./make.sh clean has run, no? > I'm just fedup to tweak make.sh to gain time will developping. Checking > all 'lfs/script' is loose time. Removing manually log/xyz is painfull, We just have not understood how to answer to some of the questions on the= menu you do. This does not mean that we should not have a menu to configure. But make.sh should be only dedicated to the build. I could send to the list my script to copy and untar pxe package to the t= ftp server, mount the iso in a loop available on http serveur, start apache a= nd tftp server. But I don't think this need to be include in make.sh. Everyone should be = free to do the way he like. In contrary, include PARRALELISM in the menu setting could be better than= hard the value in make.sh as it may be set higher on machines with multiple co= re and or HT. > copying images gz is painfull also. Checking MD5 everytime is loos time= . Some packages md5 have changed in the past without name change. This is why md5 is checked before every build to be certain the right fil= e is in the cache. I think it is a good policy to test the toolchain package validity before= to untar. More it would be better to have same test on install to check ipcop-<$VERSION>.tgz. This would detect a corrupted iso or bad write/rea= d to cdrom that happen for some people. > Think the copy is done via ssh > whith dsa key/pair for authentication (no password prompt) to > servers. yes my developping machine is not the https server. > > Every minute lost in stupid repetitive task and it is two or three > test something cycle reported at the end of the night. Write your own script to do those tasks, but this does not mean everyone = need it. > > 1.5 cycle to test 'intaller' is 59 second > 1.4 is 2minutes > I need only 5 sec after make.sh has run to be able to test a change becau= se I write my own script to do the ancillary tasks. > What is mandatory here? TWO VARIABLES for now > libc=3Dsmothing > machine=3Dsomething > We don't need to ask for MACHINE unless the toolchain support full cross-compilation (compile ppc on x86 for example) wich is not the case actually, no? This would let us in what Yvan has made with the only glibc/uclibc questi= on before. Gilles |