From: Dan A. <da...@co...> - 2004-03-13 17:44:06
|
Hello, The last version in snapshots (which I added today) contains the reimplemented allocation method. I'm delaying the release since one issue needs to be resolved with this allocation method. There is a shortcoming with Windows' cache memory policy. On my 256MB setup, Windows sometimes refuses to release cache pages that could accommodate memory for coLinux. This means that with large memory settings, it is hard to tell whether the next coLinux boot would succeed. On that machine, the 128MB setting reaches that limit, and I am forced to bring it down to somewhat 110MB for it to boot. Another issue is that I don't have machines with more than 256MB of RAM to test this version with. I'd like to know about the typical behavior on setups with bigger amounts of RAM. So, try the snapshot with various RAM settings and send results. -- Dan Aloni da...@co... |
From: Sean B. <sea...@so...> - 2004-03-14 07:26:54
|
Hi, Attempting to build the 20040313 snapshot. Using tool chain created from cobuild.sh the make fails with: i686-pc-cygwin-gcc colinux/common/common.o colinux/os/current/user/user.o colinux/os/current/user/conet-daemon/main.o colinux /os/current/user/conet-daemon/tap-win32.o colinux/user/user.a -o colinux/os/current/user/conet-daemon/colinux-net-daemon.exe -L/usr/src/colinux/cygwin/target/i686-pc-cygwin/lib/w32api -luser32 -lgdi32 -lkernel32 -lmxml i686-pc-cygwin-gcc -mpush-args -mno-accumulate-outgoing-args -Wno-trigraphs -O2 -fno-strict-aliasing -Wall -D__KERNEL__ -DCO _KERNEL -DCO_HOST_KERNEL -DWINVER=0x05000 -I. -I./../../linux/include -I/usr/src/colinux/cygwin/target/i686-pc-cygwin/include /w32api -c colinux/kernel/monitor.c -o colinux/kernel/monitor.o colinux/kernel/monitor.c: In function `colinux_init': colinux/kernel/monitor.c:140: error: `CO_VPTR_PSEUDO_RAM_PAGE_TABLES' undeclared (first use in this function) colinux/kernel/monitor.c:140: error: (Each undeclared identifier is reported only once colinux/kernel/monitor.c:140: error: for each function it appears in.) colinux/kernel/monitor.c:153: error: `CO_VPTR_PHYSICAL_TO_PSEUDO_PFN_MAP' undeclared (first use in this function) colinux/kernel/monitor.c:165: error: `CO_VPTR_SELF_MAP' undeclared (first use in this function) colinux/kernel/monitor.c:191: error: `CO_VPTR_PASSAGE_PAGE' undeclared (first use in this function) colinux/kernel/monitor.c: In function `co_monitor_iteration': colinux/kernel/monitor.c:408: error: structure has no member named `linuxvm_state' colinux/kernel/monitor.c:409: error: structure has no member named `linuxvm_state' colinux/kernel/monitor.c:410: error: structure has no member named `linuxvm_state' colinux/kernel/monitor.c:411: error: structure has no member named `linuxvm_state' colinux/kernel/monitor.c:412: error: structure has no member named `linuxvm_state' colinux/kernel/monitor.c:413: error: structure has no member named `linuxvm_state' colinux/kernel/monitor.c:414: error: structure has no member named `linuxvm_state' colinux/kernel/monitor.c: In function `alloc_pseudo_physical_memory': colinux/kernel/monitor.c:585: error: `CO_VPTR_PSEUDO_RAM_PAGE_TABLES' undeclared (first use in this function) colinux/kernel/monitor.c: In function `co_monitor_create': colinux/kernel/monitor.c:704: error: `CO_VPTR_PASSAGE_PAGE' undeclared (first use in this function) colinux/kernel/monitor.c: At top level: colinux/kernel/monitor.c:100: warning: `colinux_dump_page_at_address' defined but not used make: *** [colinux/kernel/monitor.o] Error 1 I have also created gentoo ebuilds for the toolchain as well as colinux. Using these builds the make fails with: i686-pc-cygwin-gcc -Wl,--base-file,colinux/os/current/build/driver.base.tmp \ -Wl,--entry,_DriverEntry@8 \ -nostartfiles -nostdlib \ -o junk.tmp colinux/os/current/build/driver.o -lntoskrnl -lhal -lgcc colinux/os/current/build/driver.o(.text+0x43f2):alloc.c: undefined reference to `__imp__MmAllocatePagesForMdl@28' colinux/os/current/build/driver.o(.text+0x44b9):alloc.c: undefined reference to `__imp__MmMapIoSpace@16' collect2: ld returned 1 exit status make: *** [colinux/os/current/build/driver.base.tmp] Error 1 With the environment created with cobuild.sh and my ebuilds I have been able to build all snapshots and releases up until this snapshot. Any ideas of what is happening? I think it may be possible that the tool chain needs to be updated in some way. I also think it may be to do with the linker but I have no idea and as far as I can tell it is finding the w32api libs ie notoskrnl. |
From: Thomas F. <tf...@no...> - 2004-03-14 11:33:21
|
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Hi, I don't have system access now to check this out in detail, but it looks like a problem in colinux. Also the name for the symbols looks like the parts of the program where Dan is currently working. Greetings, Thomas > Hi, > > Attempting to build the 20040313 snapshot. > > Using tool chain created from cobuild.sh the make fails with: > > i686-pc-cygwin-gcc colinux/common/common.o > colinux/os/current/user/user.o > colinux/os/current/user/conet-daemon/main.o colinux > /os/current/user/conet-daemon/tap-win32.o colinux/user/user.a -o > colinux/os/current/user/conet-daemon/colinux-net-daemon.exe > -L/usr/src/colinux/cygwin/target/i686-pc-cygwin/lib/w32api -luser32 > -lgdi32 -lkernel32 -lmxml > i686-pc-cygwin-gcc -mpush-args -mno-accumulate-outgoing-args > -Wno-trigraphs -O2 -fno-strict-aliasing -Wall -D__KERNEL__ -DCO > _KERNEL -DCO_HOST_KERNEL -DWINVER=0x05000 -I. -I./../../linux/include > -I/usr/src/colinux/cygwin/target/i686-pc-cygwin/include > /w32api -c colinux/kernel/monitor.c -o colinux/kernel/monitor.o > colinux/kernel/monitor.c: In function `colinux_init': > colinux/kernel/monitor.c:140: error: `CO_VPTR_PSEUDO_RAM_PAGE_TABLES' > undeclared (first use in this function) > colinux/kernel/monitor.c:140: error: (Each undeclared identifier is > reported only once > colinux/kernel/monitor.c:140: error: for each function it appears in.) > colinux/kernel/monitor.c:153: error: > `CO_VPTR_PHYSICAL_TO_PSEUDO_PFN_MAP' undeclared (first use in this > function) > colinux/kernel/monitor.c:165: error: `CO_VPTR_SELF_MAP' undeclared > (first use in this function) > colinux/kernel/monitor.c:191: error: `CO_VPTR_PASSAGE_PAGE' undeclared > (first use in this function) > colinux/kernel/monitor.c: In function `co_monitor_iteration': > colinux/kernel/monitor.c:408: error: structure has no member named > `linuxvm_state' > colinux/kernel/monitor.c:409: error: structure has no member named > `linuxvm_state' > colinux/kernel/monitor.c:410: error: structure has no member named > `linuxvm_state' > colinux/kernel/monitor.c:411: error: structure has no member named > `linuxvm_state' > colinux/kernel/monitor.c:412: error: structure has no member named > `linuxvm_state' > colinux/kernel/monitor.c:413: error: structure has no member named > `linuxvm_state' > colinux/kernel/monitor.c:414: error: structure has no member named > `linuxvm_state' > colinux/kernel/monitor.c: In function `alloc_pseudo_physical_memory': > colinux/kernel/monitor.c:585: error: `CO_VPTR_PSEUDO_RAM_PAGE_TABLES' > undeclared (first use in this function) > colinux/kernel/monitor.c: In function `co_monitor_create': > colinux/kernel/monitor.c:704: error: `CO_VPTR_PASSAGE_PAGE' undeclared > (first use in this function) > colinux/kernel/monitor.c: At top level: > colinux/kernel/monitor.c:100: warning: `colinux_dump_page_at_address' > defined but not used > make: *** [colinux/kernel/monitor.o] Error 1 > > I have also created gentoo ebuilds for the toolchain as well as colinux. > Using these builds the make fails with: > > i686-pc-cygwin-gcc > -Wl,--base-file,colinux/os/current/build/driver.base.tmp \ > -Wl,--entry,_DriverEntry@8 \ > -nostartfiles -nostdlib \ > -o junk.tmp colinux/os/current/build/driver.o -lntoskrnl -lhal > -lgcc > colinux/os/current/build/driver.o(.text+0x43f2):alloc.c: undefined > reference to `__imp__MmAllocatePagesForMdl@28' > colinux/os/current/build/driver.o(.text+0x44b9):alloc.c: undefined > reference to `__imp__MmMapIoSpace@16' > collect2: ld returned 1 exit status > make: *** [colinux/os/current/build/driver.base.tmp] Error 1 > > With the environment created with cobuild.sh and my ebuilds I have been > able to build all snapshots and releases up until this snapshot. Any > ideas of what is happening? > I think it may be possible that the tool chain needs to be updated in > some way. I also think it may be to do with the linker but I have no > idea and as far as I can tell it is finding the w32api libs ie > notoskrnl. > > > > > ------------------------------------------------------- > This SF.Net email is sponsored by: IBM Linux Tutorials > Free Linux tutorial presented by Daniel Robbins, President and CEO of > GenToo technologies. Learn everything from fundamentals to system > administration.http://ads.osdn.com/?ad_id=1470&alloc_id=3638&op=click > _______________________________________________ > coLinux-devel mailing list > coL...@li... > https://lists.sourceforge.net/lists/listinfo/colinux-devel > -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.2.4 (GNU/Linux) iD8DBQFAVENFnwJBIFTVIqwRAvrLAJwOefATrqqjNDf3gjShy760OvOyWQCffsF3 Q791vzal7UhDyLpkSYDAW0A= =wRCj -----END PGP SIGNATURE----- |
From: Sean B. <sea...@so...> - 2004-03-14 16:53:47
|
Hello, Thank you very much for that information. While I am at it, thanks for cobuild.sh too :) Cheers, Sean. > -----Original Message----- > From: col...@li... > [mailto:col...@li...] On Behalf > Of Thomas Fritzsche > Sent: Sunday, March 14, 2004 11:35 AM > To: col...@li... > Subject: Re: [coLinux-devel] building 20040313/cannot build > > > -----BEGIN PGP SIGNED MESSAGE----- > Hash: SHA1 > > Hi, > > I don't have system access now to check this out in detail, > but it looks like a problem in colinux. Also the name for the > symbols looks like the parts of the program where Dan is > currently working. > > Greetings, Thomas > > > > Hi, > > > > Attempting to build the 20040313 snapshot. > > > > Using tool chain created from cobuild.sh the make fails with: > > > > i686-pc-cygwin-gcc colinux/common/common.o > > colinux/os/current/user/user.o > > colinux/os/current/user/conet-daemon/main.o colinux > > /os/current/user/conet-daemon/tap-win32.o colinux/user/user.a -o > > colinux/os/current/user/conet-daemon/colinux-net-daemon.exe > > -L/usr/src/colinux/cygwin/target/i686-pc-cygwin/lib/w32api -luser32 > > -lgdi32 -lkernel32 -lmxml i686-pc-cygwin-gcc -mpush-args > > -mno-accumulate-outgoing-args -Wno-trigraphs -O2 > -fno-strict-aliasing > > -Wall -D__KERNEL__ -DCO _KERNEL -DCO_HOST_KERNEL > -DWINVER=0x05000 -I. > > -I./../../linux/include > > -I/usr/src/colinux/cygwin/target/i686-pc-cygwin/include > > /w32api -c colinux/kernel/monitor.c -o colinux/kernel/monitor.o > > colinux/kernel/monitor.c: In function `colinux_init': > > colinux/kernel/monitor.c:140: error: > `CO_VPTR_PSEUDO_RAM_PAGE_TABLES' > > undeclared (first use in this function) > > colinux/kernel/monitor.c:140: error: (Each undeclared identifier is > > reported only once > > colinux/kernel/monitor.c:140: error: for each function it > appears in.) > > colinux/kernel/monitor.c:153: error: > > `CO_VPTR_PHYSICAL_TO_PSEUDO_PFN_MAP' undeclared (first use in this > > function) > > colinux/kernel/monitor.c:165: error: `CO_VPTR_SELF_MAP' undeclared > > (first use in this function) > > colinux/kernel/monitor.c:191: error: `CO_VPTR_PASSAGE_PAGE' > undeclared > > (first use in this function) > > colinux/kernel/monitor.c: In function `co_monitor_iteration': > > colinux/kernel/monitor.c:408: error: structure has no member named > > `linuxvm_state' > > colinux/kernel/monitor.c:409: error: structure has no member named > > `linuxvm_state' > > colinux/kernel/monitor.c:410: error: structure has no member named > > `linuxvm_state' > > colinux/kernel/monitor.c:411: error: structure has no member named > > `linuxvm_state' > > colinux/kernel/monitor.c:412: error: structure has no member named > > `linuxvm_state' > > colinux/kernel/monitor.c:413: error: structure has no member named > > `linuxvm_state' > > colinux/kernel/monitor.c:414: error: structure has no member named > > `linuxvm_state' > > colinux/kernel/monitor.c: In function > `alloc_pseudo_physical_memory': > > colinux/kernel/monitor.c:585: error: > `CO_VPTR_PSEUDO_RAM_PAGE_TABLES' > > undeclared (first use in this function) > > colinux/kernel/monitor.c: In function `co_monitor_create': > > colinux/kernel/monitor.c:704: error: `CO_VPTR_PASSAGE_PAGE' > undeclared > > (first use in this function) > > colinux/kernel/monitor.c: At top level: > > colinux/kernel/monitor.c:100: warning: > `colinux_dump_page_at_address' > > defined but not used > > make: *** [colinux/kernel/monitor.o] Error 1 > > > > I have also created gentoo ebuilds for the toolchain as well as > > colinux. Using these builds the make fails with: > > > > i686-pc-cygwin-gcc > > -Wl,--base-file,colinux/os/current/build/driver.base.tmp \ > > -Wl,--entry,_DriverEntry@8 \ -nostartfiles -nostdlib \ > > -o junk.tmp colinux/os/current/build/driver.o > -lntoskrnl -lhal > > -lgcc > > colinux/os/current/build/driver.o(.text+0x43f2):alloc.c: undefined > > reference to `__imp__MmAllocatePagesForMdl@28' > > colinux/os/current/build/driver.o(.text+0x44b9):alloc.c: undefined > > reference to `__imp__MmMapIoSpace@16' > > collect2: ld returned 1 exit status > > make: *** [colinux/os/current/build/driver.base.tmp] Error 1 > > > > With the environment created with cobuild.sh and my ebuilds I have > > been able to build all snapshots and releases up until this > snapshot. > > Any ideas of what is happening? I think it may be possible that the > > tool chain needs to be updated in some way. I also think it > may be to > > do with the linker but I have no idea and as far as I can > tell it is > > finding the w32api libs ie notoskrnl. > > > > > > > > > > ------------------------------------------------------- > > This SF.Net email is sponsored by: IBM Linux Tutorials > > Free Linux tutorial presented by Daniel Robbins, President > and CEO of > > GenToo technologies. Learn everything from fundamentals to system > > > administration.http://ads.osdn.com/?ad_id=1470&alloc_id=3638&op=click > > _______________________________________________ > > coLinux-devel mailing list coL...@li... > > https://lists.sourceforge.net/lists/listinfo/colinux-devel > > > > > -----BEGIN PGP SIGNATURE----- > Version: GnuPG v1.2.4 (GNU/Linux) > > iD8DBQFAVENFnwJBIFTVIqwRAvrLAJwOefATrqqjNDf3gjShy760OvOyWQCffsF3 > Q791vzal7UhDyLpkSYDAW0A= > =wRCj > -----END PGP SIGNATURE----- > > > ------------------------------------------------------- > This SF.Net email is sponsored by: IBM Linux Tutorials > Free Linux tutorial presented by Daniel Robbins, President > and CEO of GenToo technologies. Learn everything from > fundamentals to system > administration.http://ads.osdn.com/?ad_id=1470&alloc_id=3638&op=click > _______________________________________________ > coLinux-devel mailing list > coL...@li... > https://lists.sourceforge.net/lists/listinfo/colinux-devel > |
From: Dan A. <da...@co...> - 2004-03-14 16:57:30
|
On Sun, Mar 14, 2004 at 06:31:33AM -0000, Sean Brook wrote: > Hi, > > Attempting to build the 20040313 snapshot. It looks like you forgot to apply the new version of linux patch. Plus, you need to recompile a patched version of w32api from sources, with fixed declarations for the two newly imported functions (look in ntoskrnl.def). -- Dan Aloni da...@co... |
From: Sean B. <sea...@so...> - 2004-03-14 20:13:44
Attachments:
w32api-2.4.patch
|
On Sunday 14 March 2004 16:57, Dan Aloni wrote: > On Sun, Mar 14, 2004 at 06:31:33AM -0000, Sean Brook wrote: > > Hi, > > > > Attempting to build the 20040313 snapshot. > > It looks like you forgot to apply the new version of linux patch. Yes I forgot to apply the new version of the patch to the kernel tree I use with cobuld.sh My gentoo ebuilds automatically patch a vanilla source tarball which is completely separate from the tree I use with cobuild.sh > > Plus, you need to recompile a patched version of w32api from > sources, with fixed declarations for the two newly imported > functions (look in ntoskrnl.def). Great stuff. Thanks for the pointer. I made a new patch and it compiled no problems. I made a new w32api-2.4.patch and attached it to this mail. Any advice on where to put the winpcap dev kit so I can build conet-bridged-daemon? Cheers, Sean. |
From: Alejandro R. S. <as...@MI...> - 2004-03-14 21:19:08
|
I've been working on this on and off, will get to it more once my laptop gets back from Dell. My build env has it i686-pc-cygwin/include, folded in with everything else there. It's really a hack. What I'd like to see is it in i686-pc-cygwin/include/wpcap or the like, I just don't have the make/gcc foo to get that added to the include path. Alternativly, and the option I'm leaning more towards now, is to patch the pcap include files so they know they're in wpcap/, and deal appropriatly. Things to note: the wpcap sdk looks like it was meant for windows, the zip file has case issues in three files if I recall correctly. Also, there's a types file that redefines some types (bittypes.h), which should be replaced with a symlink to the cygwin stdint.h file. -Alejandro On Sun, 2004-03-14 at 06:09, Sean Brook wrote: > On Sunday 14 March 2004 16:57, Dan Aloni wrote: > > On Sun, Mar 14, 2004 at 06:31:33AM -0000, Sean Brook wrote: > > > Hi, > > > > > > Attempting to build the 20040313 snapshot. > > > > It looks like you forgot to apply the new version of linux patch. > Yes I forgot to apply the new version of the patch to the kernel > tree I use with cobuld.sh My gentoo ebuilds automatically patch > a vanilla source tarball which is completely separate from the > tree I use with cobuild.sh > > > > > Plus, you need to recompile a patched version of w32api from > > sources, with fixed declarations for the two newly imported > > functions (look in ntoskrnl.def). > Great stuff. Thanks for the pointer. I made a new patch and it > compiled no problems. I made a new w32api-2.4.patch and > attached it to this mail. > > Any advice on where to put the winpcap dev kit so I can > build conet-bridged-daemon? > > Cheers, > > Sean. |
From: Dan A. <da...@co...> - 2004-03-14 21:19:34
|
On Sun, Mar 14, 2004 at 11:09:19AM +0000, Sean Brook wrote: > On Sunday 14 March 2004 16:57, Dan Aloni wrote: > > On Sun, Mar 14, 2004 at 06:31:33AM -0000, Sean Brook wrote: > > > Hi, > > > > > > Attempting to build the 20040313 snapshot. > > > > It looks like you forgot to apply the new version of linux patch. > Yes I forgot to apply the new version of the patch to the kernel > tree I use with cobuld.sh My gentoo ebuilds automatically patch > a vanilla source tarball which is completely separate from the > tree I use with cobuild.sh > > > > > Plus, you need to recompile a patched version of w32api from > > sources, with fixed declarations for the two newly imported > > functions (look in ntoskrnl.def). > Great stuff. Thanks for the pointer. I made a new patch and it > compiled no problems. I made a new w32api-2.4.patch and > attached it to this mail. Cool, I was about to produce the same patch. I forwarded this to the author of w32api. > Any advice on where to put the winpcap dev kit so I can > build conet-bridged-daemon? The files should be put in your $PREFIX/i686-pc-cygwin/lib and $PREFIX/i686-pc-cygwin/include directories. -- Dan Aloni da...@co... |
From: Sean B. <sea...@so...> - 2004-03-14 21:35:12
|
> > Any advice on where to put the winpcap dev kit so I can > > build conet-bridged-daemon? > > The files should be put in your $PREFIX/i686-pc-cygwin/lib and > $PREFIX/i686-pc-cygwin/include directories. Yes I presumed as much but the last time I looked there were name collisions with headers and definitions. I want to know how you did it. |
From: Thomas F. <tf...@no...> - 2004-03-15 09:23:45
|
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Hi Dan, could you please post this w32api-2.4.patch to the list (or wiki or in the next release). Thx, Thomas >> > Plus, you need to recompile a patched version of w32api from >> > sources, with fixed declarations for the two newly imported >> > functions (look in ntoskrnl.def). >> Great stuff. Thanks for the pointer. I made a new patch and it >> compiled no problems. I made a new w32api-2.4.patch and >> attached it to this mail. > > Cool, I was about to produce the same patch. I forwarded this to the > author of w32api. > -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.2.4 (GNU/Linux) iD8DBQFAVXZxnwJBIFTVIqwRAoRBAJ99B1QkN/grMDukO29KyPL994Zp1QCdE/Iv 3fUerevZg0304E1Zhl7TSLw= =ldJM -----END PGP SIGNATURE----- |
From: Clemmitt M. S. <sig...@bl...> - 2004-03-15 21:32:42
|
Hi all, I've read the building and cygwin-cross-build files in the source tarball's doc directory in preparation for trying out a build. But I have some *really* dumb questions, if anyone would be willing to answer them: 1.) I couldn't find this in the building file but, if I understand correctly, the build system for both the coLinux kernel (vmlinux) and the OS daemons and support code is i686-pc-linux. IOW, you build everything on a running Linux system, not under Windows. Is this correct? 2.) Is coLinux self-hosting, that is, can I build coLinux on a running (and stable) coLinux system? 3.) Is it dumb to ask if it's possible to use MinGW rather than Cygwin to do builds? I know about using Cygwin and MinGW on Windows as a shell and build environment, but I don't have any experience using them as a build tool platform under Linux. I do at least see that w32api from MinGW is being used.... If I have time to work on building, I'll add the info I learn along the way to the Wiki. TIA. Clemmitt Sigler |
From: Dmitriy K. <dm...@ka...> - 2004-03-15 23:04:45
|
Welcome, Clemmitt. You wrote 16 Mar 2004 ., 3:32:32: CMS> 1.) I couldn't find this in the building file but, if I understand CMS> correctly, the build system for both the coLinux kernel (vmlinux) an= d CMS> the OS daemons and support code is i686-pc-linux. IOW, you build As I understood, OS daemons/support code are built using i686-pc-cygwin CMS> everything on a running Linux system, not under Windows. Is this CMS> correct? no CMS> 2.) Is coLinux self-hosting, that is, can I build coLinux on a CMS> running (and stable) coLinux system? I did it sucessfully with 0.5.4 (I used Debian 1gb image,updated from Debian Woody 3.r1 CDs and heavily patched cobuild.sh) (but you need modern(2.95 wasn't worked for me) GCC/binutils installed FI= RST and build cross-development CYGWIN toolchain after than --=20 =D1 =F3=E2=E0=E6=E5=ED=E8=E5=EC, Dmitriy mailto:dm...@ka... |
From: Sean B. <sea...@so...> - 2004-03-16 03:30:22
|
> -----Original Message----- > From: col...@li... > [mailto:col...@li...] On Behalf > Of Clemmitt M. Sigler > Sent: Monday, March 15, 2004 9:33 PM > To: col...@li... > Subject: [coLinux-devel] Getting started building coLinux. > > > Hi all, > > I've read the building and cygwin-cross-build files in the > source tarball's doc directory in preparation for trying out > a build. But I have some *really* dumb questions, if anyone > would be willing to answer them: > > 1.) I couldn't find this in the building file but, if I understand > correctly, the build system for both the coLinux kernel > (vmlinux) and the OS daemons and support code is > i686-pc-linux. IOW, you build everything on a running Linux > system, not under Windows. Is this correct? I would say this is correct. The build host is ???-??-linux. Creating the cygwin tool chain enables cross compiling to a a i686-pc-cygwin target > > 2.) Is coLinux self-hosting, that is, can I build colinux on > a running (and stable) coLinux system? Yes, and I think that should be your intention. colinux is stable enough to do that (use a binary to begin?). IMO it goes to avoid some pitfalls like using different versions of gcc and compiling modules inside colinux with headers different from that which you used to compile the kernel. > > 3.) Is it dumb to ask if it's possible to use MinGW rather > than Cygwin to do builds? I know about using Cygwin and > MinGW on Windows as a shell and build environment, but I > don't have any experience using them as a build tool platform > under Linux. I do at least see that w32api from MinGW is > being used.... This might be interesting. I don't know why though. IMO it is best to keep the build environment the same as the devs. > > If I have time to work on building, I'll add the info I learn > along the way to the Wiki. TIA. > > Clemmitt Sigler > > > > ------------------------------------------------------- > This SF.Net email is sponsored by: IBM Linux Tutorials > Free Linux tutorial presented by Daniel Robbins, President > and CEO of GenToo technologies. Learn everything from > fundamentals to system > administration.http://ads.osdn.com/?ad_id=1470&alloc_id=3638&op=click > _______________________________________________ > coLinux-devel mailing list > coL...@li... > https://lists.sourceforge.net/lists/listinfo/colinux-devel > |
From: <ch...@to...> - 2004-03-16 12:52:03
|
>> 3.) Is it dumb to ask if it's possible to use MinGW rather >> than Cygwin to do builds? I know about using Cygwin and >> MinGW on Windows as a shell and build environment, but I >> don't have any experience using them as a build tool platform >> under Linux. I do at least see that w32api from MinGW is >> being used.... > This might be interesting. I don't know why though. IMO it is > best to keep the build environment the same as the devs. > To the best of my knowlege. If the cygwin1.dll is needed for the posix layer it provides then It would not be possible to build under MinGW. MinGW is just a port of GCC to windows for programing with the windows api Cygwin provides a posix layer for easy porting of *NIX programs to windows. If the posix layer is not needed then MinGW would be the better build environment so that the cygwin1.dll is not required. The cygwin1.dll is very sensitive to having multiple versions on machines. I do not know if backwards compatibility is always maintained so it may not always be as simple as replacing cygwin1.dll in either coLinux or the cygwin install with the newer one. chris |
From: tei <42...@in...> - 2004-03-16 18:04:00
|
ch...@to... escribió: >>>3.) Is it dumb to ask if it's possible to use MinGW rather >>>than Cygwin to do builds? I know about using Cygwin and >>>MinGW on Windows as a shell and build environment, but I >>>don't have any experience using them as a build tool platform >>>under Linux. I do at least see that w32api from MinGW is >>>being used.... >> >>This might be interesting. I don't know why though. IMO it is >>best to keep the build environment the same as the devs. >> > > > To the best of my knowlege. If the cygwin1.dll is needed for the posix > layer it provides then It would not be possible to build under MinGW. > MinGW is just a port of GCC to windows for programing with the windows api > Cygwin provides a posix layer for easy porting of *NIX programs to > windows. > > If the posix layer is not needed then MinGW would be the better build > environment so that the cygwin1.dll is not required. The cygwin1.dll is > very sensitive to having multiple versions on machines. I do not know if > backwards compatibility is always maintained so it may not always be as > simple as replacing cygwin1.dll in either coLinux or the cygwin install > with the newer one. > > chris > Hi! <offtopic> <info about cygwin> The need for cygwin1.dll is optional. You can build the same binary from MinGW and Cygwin. Except.. Cygwin provice some extra features on Cygwin1.dll, of course, If you use that features, you need that dll. So, MinGW is cool (the Min means minimal?), and Cygwin is cool. Both are cool. note: You may already know that, this is a reminder for others. </info about cygwin> </offtopic> |
From: Ballard J. <sac...@ho...> - 2004-03-16 20:05:03
|
> >>>3.) Is it dumb to ask if it's possible to use MinGW rather > >>>than Cygwin to do builds? > The need for cygwin1.dll is optional. It is possible to use only MinGW. Currently, the source code uses standard C (stdlib, stdio, etc) conventions and Microsoft API's only have limited implementation of such conventions. Microsoft has an optional run-time library, installed on demand like cygwin1.dll, that could be used intead of Cygwin or MinGW but the benefits does not change because an extra *.dll file was required. The best option is to rewrite the portions of the coLinux source to use only the native DDK and PlatformSDK conventions and procedures. I do not see such change as a priority issue at this time. As coLinux becomes more secure, I project the rewrite to more native OS code to happen. The standard C conventions are easier use at this time -- especially for those that program with linux more. |
From: Steven E. <ste...@ya...> - 2004-03-16 22:04:32
|
--- Ballard Jonathan <sac...@ho...> wrote: > It is possible to use only MinGW. Currently, the source code uses > standard > C (stdlib, stdio, etc) conventions and Microsoft API's only have > limited > implementation of such conventions. Microsoft has an optional > run-time > library, installed on demand like cygwin1.dll, that could be used > intead of > Cygwin or MinGW but the benefits does not change because an extra > *.dll file > was required. The best option is to rewrite the portions of the > coLinux > source to use only the native DDK and PlatformSDK conventions and > procedures. I do not see such change as a priority issue at this > time. As > coLinux becomes more secure, I project the rewrite to more native OS > code to > happen. The standard C conventions are easier use at this time -- > especially for those that program with linux more. You could do it from a Mingw-Cross compiler on Linux. I think one of our developers was able to build CoLinux using the Mingw linker rather than linking to Cygwin. Thanks Steven __________________________________ Do you Yahoo!? Yahoo! Mail - More reliable, more storage, less spam http://mail.yahoo.com |
From: Ballard J. <sac...@ho...> - 2004-03-17 00:14:36
|
Subject: Re: [coLinux-devel] Getting started building coLinux. > --- Ballard Jonathan <sac...@ho...> wrote: > > It is possible to use only MinGW. Currently, the source code uses [...] > > You could do it from a Mingw-Cross compiler on Linux. I think one of > our developers was able to build CoLinux using the Mingw linker rather > than linking to Cygwin. I heavily modified and compiled coLinux sources, headers, and a portion of the linux system without GCC, cygwin or MinGW. I used Microsoft Development Environment (MDE). I reached a point where some linux sources were specifically written for GCC style compilers and that made it very tedious. I didn't have the time to rewrite the sources to be more portable. I know it is possible to compile without cygwin. I think the 2004 version of MDE, with updates, would make it a bit easier since it includes some of the recent C standards. For a less expensive immediate alternative, I suggest to grab a coLinux binary and use GCC and either cygwin or MinGW. |