| Message ID | 296699691239a90d5118c5cbe72f1ff57b3761a5.1362693294.git.yann.morin.1998@free.fr |
|---|---|
| State | Changes Requested |
| Headers | show |
Yann, Luca, Cc'ing Luca, there is a licensing infrastructure question below. On Thu, 7 Mar 2013 22:55:31 +0100, Yann E. MORIN wrote: > +MTOOLS_VERSION = 4.0.18 > +MTOOLS_SOURCE = mtools-$(MTOOLS_VERSION).tar.bz2 > +MTOOLS_SITE = $(BR2_GNU_MIRROR)/mtools/ MTOOLS_LICENSE = GPLv3+ MTOOLS_LICENSE_FILES = COPYING However, this makes me realize that we don't distinguish host packages from target packages in the manifest.csv file. For example, once I had the licensing informations for this mtools file, I have the following line in my manifest.csv: "mtools","4.0.18","GPLv3+","COPYING","mtools-4.0.18.tar.bz2" And then, someone might think "damn, I have some GPLv3 software in my system, I now need to comply with the additional special requirements of GPLv3", while in fact it's not the case (as far as I know) because mtools is not distributed on the embedded device. Maybe we need to explicit which packages are on the target, which packages are on the host, and which packages are on both, no? Best regards, Thomas
Thomas, Luca, All, On Sunday 10 March 2013 Thomas Petazzoni wrote: > On Thu, 7 Mar 2013 22:55:31 +0100, Yann E. MORIN wrote: > > > +MTOOLS_VERSION = 4.0.18 > > +MTOOLS_SOURCE = mtools-$(MTOOLS_VERSION).tar.bz2 > > +MTOOLS_SITE = $(BR2_GNU_MIRROR)/mtools/ > > MTOOLS_LICENSE = GPLv3+ > MTOOLS_LICENSE_FILES = COPYING Done. > However, this makes me realize that we don't distinguish host packages > from target packages in the manifest.csv file. For example, once I had > the licensing informations for this mtools file, I have the following > line in my manifest.csv: > > "mtools","4.0.18","GPLv3+","COPYING","mtools-4.0.18.tar.bz2" > > And then, someone might think "damn, I have some GPLv3 software in my > system, I now need to comply with the additional special requirements > of GPLv3", while in fact it's not the case (as far as I know) because > mtools is not distributed on the embedded device. Maybe we need to > explicit which packages are on the target, which packages are on the > host, and which packages are on both, no? Yes, fully agreed; this would be very puzzling. I would suggest that we not install licensing information about host packages, unless they are also used on the target. Regards, Yann E. MORIN.
Dear Yann E. MORIN, On Mon, 11 Mar 2013 01:02:06 +0100, Yann E. MORIN wrote: > I would suggest that we not install licensing information about host > packages, unless they are also used on the target. That would make sense, IMO. Thomas
Yann E. MORIN wrote: > Thomas, Luca, All, > > On Sunday 10 March 2013 Thomas Petazzoni wrote: >> On Thu, 7 Mar 2013 22:55:31 +0100, Yann E. MORIN wrote: >> >>> +MTOOLS_VERSION = 4.0.18 >>> +MTOOLS_SOURCE = mtools-$(MTOOLS_VERSION).tar.bz2 >>> +MTOOLS_SITE = $(BR2_GNU_MIRROR)/mtools/ >> >> MTOOLS_LICENSE = GPLv3+ >> MTOOLS_LICENSE_FILES = COPYING > > Done. > >> However, this makes me realize that we don't distinguish host packages >> from target packages in the manifest.csv file. For example, once I had >> the licensing informations for this mtools file, I have the following >> line in my manifest.csv: >> >> "mtools","4.0.18","GPLv3+","COPYING","mtools-4.0.18.tar.bz2" >> >> And then, someone might think "damn, I have some GPLv3 software in my >> system, I now need to comply with the additional special requirements >> of GPLv3", while in fact it's not the case (as far as I know) because >> mtools is not distributed on the embedded device. Maybe we need to >> explicit which packages are on the target, which packages are on the >> host, and which packages are on both, no? > > Yes, fully agreed; this would be very puzzling. > > I would suggest that we not install licensing information about host > packages, unless they are also used on the target. It would probably be sound in most cases, but not compliant to some licenses, at least if there isa GPL package on the target. Take what the GPLv2 saysin clause 3: > The source code for a work means the preferred form of the work for > making modifications to it. For an executable work, complete source > code means all the source code for all modules it contains, plus any > associated interface definition files, plus the scripts used to > control compilation and installation of the executable. What do the "scripts used to control compilation and installation"? We use a toolchain, the autotools, CMake, Python etc on the host to build packages. While these tools are arguably more complex than the average "script", they are "used to control the cpmpilation and installation". Reading on: > However, as a > special exception, the source code distributed need not include > anything that is normally distributed (in either source or binary > form) with the major components (compiler, kernel, and so on) of the > operating system on which the executable runs, unless that component > itself accompanies the executable. While this is saying you do not need to redistribute the Debian compiler, it is using "compiler" and "kernel" as two examples of the "scripts used to control compilation and installation". Hence my understanding that the toolchain should be redistributed, along with all of the host tools that are used to build anything on the target. As far as the mtools are concerned, they are probably not used to produce the executables on the target, so they actually do not need to be in the redistribution list. But most other host packages are different. Indeed, in the practice I'm retty sure almost nobody modify autotools, or CMake or other host packages, so even redistributing them would not disclose much industrial secret. Bottom line, I think it would be useful to save in the package manifest an additional column stating whether the package is on the target, on the host, or both. But skipping host-only packages is not safe. Luca
Hi, On Tue, Mar 12, 2013 at 6:16 PM, Luca Ceresoli <luca@lucaceresoli.net> wrote: > Yann E. MORIN wrote: >> Thomas, Luca, All, >> >> On Sunday 10 March 2013 Thomas Petazzoni wrote: >>> On Thu, 7 Mar 2013 22:55:31 +0100, Yann E. MORIN wrote: >>> >>>> +MTOOLS_VERSION = 4.0.18 >>>> +MTOOLS_SOURCE = mtools-$(MTOOLS_VERSION).tar.bz2 >>>> +MTOOLS_SITE = $(BR2_GNU_MIRROR)/mtools/ >>> >>> MTOOLS_LICENSE = GPLv3+ >>> MTOOLS_LICENSE_FILES = COPYING >> >> Done. >> >>> However, this makes me realize that we don't distinguish host packages >>> from target packages in the manifest.csv file. For example, once I had >>> the licensing informations for this mtools file, I have the following >>> line in my manifest.csv: >>> >>> "mtools","4.0.18","GPLv3+","COPYING","mtools-4.0.18.tar.bz2" >>> >>> And then, someone might think "damn, I have some GPLv3 software in my >>> system, I now need to comply with the additional special requirements >>> of GPLv3", while in fact it's not the case (as far as I know) because >>> mtools is not distributed on the embedded device. Maybe we need to >>> explicit which packages are on the target, which packages are on the >>> host, and which packages are on both, no? >> >> Yes, fully agreed; this would be very puzzling. >> >> I would suggest that we not install licensing information about host >> packages, unless they are also used on the target. > > It would probably be sound in most cases, but not compliant to some > licenses, at least if there isa GPL package on the target. > Take what the GPLv2 saysin clause 3: > >> The source code for a work means the preferred form of the work for >> making modifications to it. For an executable work, complete source >> code means all the source code for all modules it contains, plus any >> associated interface definition files, plus the scripts used to >> control compilation and installation of the executable. > > What do the "scripts used to control compilation and installation"? > We use a toolchain, the autotools, CMake, Python etc on the host to build > packages. While these tools are arguably more complex than the average > "script", they are "used to control the cpmpilation and installation". > > Reading on: > >> However, as a >> special exception, the source code distributed need not include >> anything that is normally distributed (in either source or binary >> form) with the major components (compiler, kernel, and so on) of the >> operating system on which the executable runs, unless that component >> itself accompanies the executable. > > While this is saying you do not need to redistribute the Debian compiler, > it is using "compiler" and "kernel" as two examples of the "scripts used > to control compilation and installation". Hence my understanding that the > toolchain should be redistributed, along with all of the host tools that > are used to build anything on the target. > > As far as the mtools are concerned, they are probably not used to produce > the executables on the target, so they actually do not need to be in the > redistribution list. But most other host packages are different. > > Indeed, in the practice I'm retty sure almost nobody modify autotools, > or CMake or other host packages, so even redistributing them would not > disclose much industrial secret. > > Bottom line, I think it would be useful to save in the package manifest > an additional column stating whether the package is on the target, on > the host, or both. But skipping host-only packages is not safe. I follow Luca's reasoning here. As we discussed on the Buildroot Developer Day 2012, we will include the toolchain/host sources precisely for the above reasons, and so they should be described in a manifest. I also agree with Thomas' point about the confusion. But, instead of an extra column, wouldn't it be more clear to separate the legal-license manifest files for host and target packages? This would result in a manifest-host.csv and manifest.csv. Packages that appear on both host and target, would be present in both files. An end user that is only interested in getting a list of target packages, can find that info in a glance. Best regards, Thomas
On 03/12/13 18:40, Thomas De Schampheleire wrote: > Hi, > > On Tue, Mar 12, 2013 at 6:16 PM, Luca Ceresoli <luca@lucaceresoli.net> wrote: [snip] >> Bottom line, I think it would be useful to save in the package manifest >> an additional column stating whether the package is on the target, on >> the host, or both. But skipping host-only packages is not safe. I agree. And I have a second reason for that: I have one customer who distributes an "SDK", i.e. host packages. > I follow Luca's reasoning here. As we discussed on the Buildroot > Developer Day 2012, we will include the toolchain/host sources > precisely for the above reasons, and so they should be described in a > manifest. I also agree with Thomas' point about the confusion. > > But, instead of an extra column, wouldn't it be more clear to separate > the legal-license manifest files for host and target packages? This > would result in a manifest-host.csv and manifest.csv. Packages that > appear on both host and target, would be present in both files. An end > user that is only interested in getting a list of target packages, can > find that info in a glance. Very good idea. Regards, Arnout
Dear Arnout Vandecappelle, On Wed, 13 Mar 2013 00:24:31 +0100, Arnout Vandecappelle wrote: > > But, instead of an extra column, wouldn't it be more clear to separate > > the legal-license manifest files for host and target packages? This > > would result in a manifest-host.csv and manifest.csv. Packages that > > appear on both host and target, would be present in both files. An end > > user that is only interested in getting a list of target packages, can > > find that info in a glance. > > Very good idea. I don't know if creating a separate file is needed. Maybe just a need field in the CSV that says if it's a host package only, target package only, or both. But that's a minor detail, I don't have a strong opinion between a single CSV file, and two CSV files, as long as it is possible to identify what is on the target and what isn't. Best regards, Thomas
diff --git a/package/Config.in.host b/package/Config.in.host index e97f6e8..7c0dd20 100644 --- a/package/Config.in.host +++ b/package/Config.in.host @@ -6,6 +6,7 @@ source "package/e2fsprogs/Config.in.host" source "package/genimage/Config.in.host" source "package/genpart/Config.in.host" source "package/lpc3250loader/Config.in.host" +source "package/mtools/Config.in.host" source "package/omap-u-boot-utils/Config.in.host" source "package/openocd/Config.in.host" source "package/sam-ba/Config.in.host" diff --git a/package/mtools/Config.in.host b/package/mtools/Config.in.host new file mode 100644 index 0000000..d534b5e --- /dev/null +++ b/package/mtools/Config.in.host @@ -0,0 +1,9 @@ +config BR2_PACKAGE_HOST_MTOOLS + bool "host mtools" + help + Mtools is a collection of utilities to access MS-DOS disks from + Unix without mounting them. It supports Win'95 style long file + names, OS/2 Xdf disks and 2m disks (store up to 1992k on a high + density 3 1/2 disk). + + http://www.tux.org/pub/tux/mtools/ diff --git a/package/mtools/mtools.mk b/package/mtools/mtools.mk new file mode 100644 index 0000000..1542799 --- /dev/null +++ b/package/mtools/mtools.mk @@ -0,0 +1,15 @@ +############################################################# +# +# dosfstools +# +############################################################# + +MTOOLS_VERSION = 4.0.18 +MTOOLS_SOURCE = mtools-$(MTOOLS_VERSION).tar.bz2 +MTOOLS_SITE = $(BR2_GNU_MIRROR)/mtools/ + +MTOOLS_CONF_OPT = --enable-xdf \ + --enable-vold \ + --enable-new-vold + +$(eval $(host-autotools-package))
Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr> Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com> --- v1 -> v2 - remove extraneous spaces, use $(BR2_GNU_MIRROR) (Thomas) --- package/Config.in.host | 1 + package/mtools/Config.in.host | 9 +++++++++ package/mtools/mtools.mk | 15 +++++++++++++++ 3 files changed, 25 insertions(+), 0 deletions(-) create mode 100644 package/mtools/Config.in.host create mode 100644 package/mtools/mtools.mk