On Wed, Aug 12, 2020 at 8:41 AM Slater, Joseph <[email protected]> wrote:
>
> The symptom seems to be that sometimes, when a lot of similar tempfiles are 
> created under /tmp, a file will be created that cannot be read once the 
> original fd is closed.  Moving where these files are created might avoid some 
> underlying problem, but it also might just reduce the failure rate to 
> something so low that you never see it.
> There should probably be more investigation into what is going on, but it is 
> hard to put much effort into dealing with something that (usually) occurs 
> very rarely and is easily avoided.

this issue has never shown up on builders that I have access to, and I
would be interested to know if there are more users seeing this issue
as well.
I think we need some more reports on these lines for this change since
it will impact everyone

>
> Joe
>
> -----Original Message-----
> From: Khem Raj <[email protected]>
> Sent: Tuesday, August 11, 2020 4:04 PM
> To: Slater, Joseph <[email protected]>
> Cc: openembeded-devel <[email protected]>; MacLeod, 
> Randy <[email protected]>
> Subject: Re: [oe] [meta-oe][PATCH 1/1] mozjs: specify TMPDIR
>
> On Tue, Aug 11, 2020 at 2:02 PM Joe Slater <[email protected]> wrote:
> >
> > Although it is extremely rare the following type of error can occur
> > during configuration:
> >
> > Creating `/tmp/conftest.i53clm4z.cpp` with content:
> > ...
> > x86_64-wrs-linux-g++: error: /tmp/conftest.i53clm4z.cpp: No such file
> > or directory
> > x86_64-wrs-linux-g++: error: /tmp/conftest.i53clm4z.cpp: No such file
> > or directory
> > x86_64-wrs-linux-g++: fatal error: no input files
> >
> > So, specify a local directory as TMPDIR during configuration.
>
> Do we know the real underlying issue ? it's not clear to me from description
>
> /tmp is always mounted on linux build systems so I am assuming its hitting 
> some sort of race condition since /tmp is mounted using tmpfs it might be 
> more evident since its faster whereas when its pointing to custom dir on HDD 
> its hiding the underlying issue.
>
> IOW how is the problem manifested and how is this patch helping ?
>
> >
> > Signed-off-by: Joe Slater <[email protected]>
> > ---
> >  .../dynamic-layers/meta-python/recipes-extended/mozjs/mozjs_60.9.0.bb
> > | 4 +++-
> >  1 file changed, 3 insertions(+), 1 deletion(-)
> >
> > diff --git
> > a/meta-oe/dynamic-layers/meta-python/recipes-extended/mozjs/mozjs_60.9
> > .0.bb
> > b/meta-oe/dynamic-layers/meta-python/recipes-extended/mozjs/mozjs_60.9
> > .0.bb
> > index f13a184..629c652 100644
> > ---
> > a/meta-oe/dynamic-layers/meta-python/recipes-extended/mozjs/mozjs_60.9
> > .0.bb
> > +++ b/meta-oe/dynamic-layers/meta-python/recipes-extended/mozjs/mozjs_
> > +++ 60.9.0.bb
> > @@ -100,7 +100,9 @@ do_configure() {
> >      autoconf213 --macrodir=${STAGING_DATADIR_NATIVE}/autoconf213
> > old-configure.in > old-configure
> >
> >      cd ${B}
> > -    ${S}/js/src/configure ${EXTRA_OECONF}
> > +    # use of /tmp can causes problems on heavily loaded hosts
> > +    mkdir -p "${B}/lcl_tmp"
> > +    TMPDIR="${B}/lcl_tmp"  ${S}/js/src/configure ${EXTRA_OECONF}
> >
> >      # Make standard Makefile checks pass
> >      touch ${S}/js/src/configure
> > --
> > 2.7.4
> >
> > 
-=-=-=-=-=-=-=-=-=-=-=-
Links: You receive all messages sent to this group.

View/Reply Online (#86280): 
https://lists.openembedded.org/g/openembedded-devel/message/86280
Mute This Topic: https://lists.openembedded.org/mt/76135435/21656
Group Owner: [email protected]
Unsubscribe: https://lists.openembedded.org/g/openembedded-devel/unsub  
[[email protected]]
-=-=-=-=-=-=-=-=-=-=-=-

Reply via email to