| Message ID | 20180625014101.13177-1-ricardo.martincoski@gmail.com |
|---|---|
| State | Superseded |
| Headers | show |
| Series | [1/1] pkg-generic: fix no-check-hash for inferred site method | expand |
Hi Ricardo, On 25/06/2018 03:41, Ricardo Martincoski wrote: > Currently, when the site method is explicitly set to a SCM other than > git, the main download is correctly excluded from being hash-checked. > But when the site method is inferred from the site uri, the download > from a SCM other than git is wrongly being hash-checked. > > Fix this by moving the code that excludes SCM methods from hash-check > below the code that infers the site method. > > Signed-off-by: Ricardo Martincoski <ricardo.martincoski@gmail.com> > Cc: Luca Ceresoli <luca@lucaceresoli.net> > Cc: Maxime Hadjinlian <maxime.hadjinlian@gmail.com> > Cc: Thomas Petazzoni <thomas.petazzoni@bootlin.com> > Cc: Yann E. MORIN <yann.morin.1998@free.fr> > --- > All packages in the tree that use those site methods currently > explicitly set the site method. > $ make -s printvars VARS=%_SITE_METHOD | grep '=\(bzr\|cvs\|hg\|svn\)' > CANFESTIVAL_SITE_METHOD=hg > DVB_APPS_SITE_METHOD=hg > FIS_SITE_METHOD=svn > OPEN2300_SITE_METHOD=svn > OPENTYRIAN_SITE_METHOD=hg > PYTHON_NFC_SITE_METHOD=bzr > PYTHON_PYGAME_SITE_METHOD=hg Apologies, but I couldn't understand what this commit is supposed to fix. From your description above I understand you're fixing packages whose method is inferred (i.e. not explicitly set). But here you list packages which explicitly set it. Which one are you fixing exactly? Can you provide an example package that your patch fixes? BTW I couldn't find in Buildroot any package whose method is an SCM other than git and is not explicitly set. Thanks,
Hello, On Mon, Jun 25, 2018 at 06:01 PM, Luca Ceresoli wrote: > On 25/06/2018 03:41, Ricardo Martincoski wrote: >> Currently, when the site method is explicitly set to a SCM other than >> git, the main download is correctly excluded from being hash-checked. >> But when the site method is inferred from the site uri, the download >> from a SCM other than git is wrongly being hash-checked. [snip] >> --- >> All packages in the tree that use those site methods currently >> explicitly set the site method. >> $ make -s printvars VARS=%_SITE_METHOD | grep '=\(bzr\|cvs\|hg\|svn\)' [snip] >> PYTHON_PYGAME_SITE_METHOD=hg > > Apologies, but I couldn't understand what this commit is supposed to > fix. From your description above I understand you're fixing packages > whose method is inferred (i.e. not explicitly set). But here you list I am fixing packages whose method is inferred. > packages which explicitly set it. Sorry. My message after the commit log was misleading. Rethinking now, a better message would be: "Currently there is no package in the tree that uses inferred site method, and that is why the autobuilder didn't caught this. We had packages using inferred site method in the past, the last one was 'expect', but since they didn't have a hash file (for a license or extra-download for example) the build didn't error out." > > Which one are you fixing exactly? Can you provide an example package > that your patch fixes? There is no example in the tree right now, but they can either live in a br2-external somewhere or they can be added to the tree in the future. I don't have a real package with this problem. I stumbled on this because I am creating automated tests for the download infra. They setup a SCM remote in the docker image by calling git daemon, svnserver, ... depending on the download method and download a package. Here is one package under test [1], the test failing before this patch [2] and the test passing after this patch [3]. > > BTW I couldn't find in Buildroot any package whose method is an SCM > other than git and is not explicitly set. Correct. The last one was 'expect', it now uses a tarball instead of cvs. You can reproduce the problem I a trying to fix by doing this: $ git revert 23ab6cb162b7456925e2c3df4917cf9b45868d55 $ touch package/expect/expect.hash $ make expect-source ERROR: No hash found for expect-2014-05-02.tar.gz [1] https://gitlab.com/RicardoMartincoski/buildroot/blob/910ad20b78db2b5f238daca18c5688e3ec3c157b/support/testing/tests/download/br2-external/svn-refs/package/svn-revision-number/svn-revision-number.mk [2] https://gitlab.com/RicardoMartincoski/buildroot/-/jobs/77355189 [3] https://gitlab.com/RicardoMartincoski/buildroot/-/jobs/77355202 Regards, Ricardo
Hi Ricardo, On 26/06/2018 04:01, Ricardo Martincoski wrote: > Hello, > > On Mon, Jun 25, 2018 at 06:01 PM, Luca Ceresoli wrote: > >> On 25/06/2018 03:41, Ricardo Martincoski wrote: >>> Currently, when the site method is explicitly set to a SCM other than >>> git, the main download is correctly excluded from being hash-checked. >>> But when the site method is inferred from the site uri, the download >>> from a SCM other than git is wrongly being hash-checked. > [snip] >>> --- >>> All packages in the tree that use those site methods currently >>> explicitly set the site method. >>> $ make -s printvars VARS=%_SITE_METHOD | grep '=\(bzr\|cvs\|hg\|svn\)' > [snip] >>> PYTHON_PYGAME_SITE_METHOD=hg >> >> Apologies, but I couldn't understand what this commit is supposed to >> fix. From your description above I understand you're fixing packages >> whose method is inferred (i.e. not explicitly set). But here you list > > I am fixing packages whose method is inferred. > >> packages which explicitly set it. > > Sorry. My message after the commit log was misleading. > Rethinking now, a better message would be: > "Currently there is no package in the tree that uses inferred site method, and > that is why the autobuilder didn't caught this. We had packages using inferred > site method in the past, the last one was 'expect', but since they didn't have a > hash file (for a license or extra-download for example) the build didn't error > out." Ok, thanks for the detailed explanation. I would append the above paragraph to the existing message, just before the SoB line. If you resubmit with this change, you can add my: Reviewed-by: Luca Ceresoli <luca@lucaceresoli.net> Tested-by: Luca Ceresoli <luca@lucaceresoli.net> Bye,
diff --git a/package/pkg-generic.mk b/package/pkg-generic.mk index 501dd3de39..91b61c6de0 100644 --- a/package/pkg-generic.mk +++ b/package/pkg-generic.mk @@ -489,10 +489,6 @@ ifndef $(2)_PATCH endif endif -ifneq ($$(filter bzr cvs hg svn,$$($(2)_SITE_METHOD)),) -BR_NO_CHECK_HASH_FOR += $$($(2)_SOURCE) -endif - $(2)_ALL_DOWNLOADS = \ $$(if $$($(2)_SOURCE),$$($(2)_SITE_METHOD)+$$($(2)_SITE)/$$($(2)_SOURCE)) \ $$(foreach p,$$($(2)_PATCH) $$($(2)_EXTRA_DOWNLOADS),\ @@ -514,6 +510,10 @@ ifndef $(2)_SITE_METHOD endif endif +ifneq ($$(filter bzr cvs hg svn,$$($(2)_SITE_METHOD)),) +BR_NO_CHECK_HASH_FOR += $$($(2)_SOURCE) +endif + # Do not accept to download git submodule if not using the git method ifneq ($$($(2)_GIT_SUBMODULES),) ifneq ($$($(2)_SITE_METHOD),git)
Currently, when the site method is explicitly set to a SCM other than git, the main download is correctly excluded from being hash-checked. But when the site method is inferred from the site uri, the download from a SCM other than git is wrongly being hash-checked. Fix this by moving the code that excludes SCM methods from hash-check below the code that infers the site method. Signed-off-by: Ricardo Martincoski <ricardo.martincoski@gmail.com> Cc: Luca Ceresoli <luca@lucaceresoli.net> Cc: Maxime Hadjinlian <maxime.hadjinlian@gmail.com> Cc: Thomas Petazzoni <thomas.petazzoni@bootlin.com> Cc: Yann E. MORIN <yann.morin.1998@free.fr> --- All packages in the tree that use those site methods currently explicitly set the site method. $ make -s printvars VARS=%_SITE_METHOD | grep '=\(bzr\|cvs\|hg\|svn\)' CANFESTIVAL_SITE_METHOD=hg DVB_APPS_SITE_METHOD=hg FIS_SITE_METHOD=svn OPEN2300_SITE_METHOD=svn OPENTYRIAN_SITE_METHOD=hg PYTHON_NFC_SITE_METHOD=bzr PYTHON_PYGAME_SITE_METHOD=hg --- package/pkg-generic.mk | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-)