Where communities thrive


  • Join over 1.5M+ people
  • Join over 100K+ communities
  • Free without limits
  • Create your own community
People
Repo info
Activity
    Caleb Maclennan
    @alerque
    It should only ever be called as a | ... order-only dependency.
    i.e. We never want to mess with updating .fonts itself only the things in it. And if it exists as a symlink we shouldn't have been asked to try to make it as a directory, it just needs to exist.
    Simon Cozens
    @simoncozens
    OK. Why the heck is it erroring out, then?
    Caleb Maclennan
    @alerque
    Does [[ -h not_a_thing ]] return true or false on your bash + filesystem?
    Simon Cozens
    @simoncozens
    bash-5.0$ [[ -h .fonts ]] ; echo $?
    1
    Caleb Maclennan
    @alerque
    With .fonts being what? A dir? nothing? a link?
    Simon Cozens
    @simoncozens
    Nothing there.
    Caleb Maclennan
    @alerque
     .fonts: fontconfig.conf
    -    [[ -h .fonts ]] && exit 1
    +    [[ -h .fonts ]] && exit 1 ||:
         mkdir -p $@
    Simon Cozens
    @simoncozens
    That works.
    Caleb Maclennan
    @alerque
    Your default shell options are different. That might trip us up elsewhere too, but we can fix it.
    Simon Cozens
    @simoncozens
    OK, figures. The test returned error 1 to make, which aborted.
    Caleb Maclennan
    @alerque
    In the event fonts doesn't exist as a symlink it should be moving on to making it as a directory.
    This should also work:
     .fonts: fontconfig.conf
    -    [[ -h .fonts ]] && exit 1 ||:
    -    mkdir -p $@
    +    [[ -h .fonts ]] || mkdir -p $@
    Simon Cozens
    @simoncozens
    Works and makes me happier.
    Caleb Maclennan
    @alerque
    Added to PR. Not sure about the deps line we dropper earlier. We kind of skipped fixing that.
    Idea....
    @@ -150,7 +150,7 @@ define runsile =
         fi
     endef
    
    -patterndeps = $(DEPDIR)/$*.d all testprep
    +patterndeps = $(wildcard $(DEPDIR)/$*.d) all testprep
    
     %.pdf: %.sil $(and $(SILE_COVERAGE)$(CLEAN),force) $$(patterndeps)
         $(runsile)
    Caleb Maclennan
    @alerque
    That's going to need another step. One sec.
    Simon Cozens
    @simoncozens
    Hang on, looks like we've gone backwards anyway.
    Caleb Maclennan
    @alerque
    @@ -143,6 +143,7 @@ pages = pdfinfo $@ | awk '$$1 == "Pages:" {print $$2}' || echo 0
     silepass = $(LOCALTESTFONTS) ./sile $(SILEFLAGS) $< -o $@ && pg0=$${pg} pg=$$($(pages))
     define runsile =
         pg0=$$($(pages)) hadtoc=$$($(hastoc))
    +    mkdir -p $(DEPDIR)
         $(silepass)
         if $(hastoc); then
             $${hadtoc} || $(silepass)
    @@ -150,7 +151,7 @@ define runsile =
         fi
     endef
    
    -patterndeps = $(DEPDIR)/$*.d all testprep
    +patterndeps = $(wildcard $(DEPDIR)/$*.d) all testprep
    
     %.pdf: %.sil $(and $(SILE_COVERAGE)$(CLEAN),force) $$(patterndeps)
         $(runsile)
    Simon Cozens
    @simoncozens
    Looks like we've got it.
    Caleb Maclennan
    @alerque
    Travis seems to agree.
    Simon Cozens
    @simoncozens
    Three known-bad tests passing...
    Otherwise OS X is happy.
    Caleb Maclennan
    @alerque
    Ya same 3 here. We know there are bugs there but the tests aren't actually catching them.
    Anything else on your end after Travis gets done taking its merry time?
    Simon Cozens
    @simoncozens
    You wanted to check OS X on a github artifact (in #768)?
    Caleb Maclennan
    @alerque
    Let me attach a new one with that make stuff fixed as in the current PR.
    Uploading....
    It's a zip (make dist-zip) because that's what I can attach to the issue, but I was planning on only posting the *.tar.bz2 source for now until we have a release on which Windows is expected to build out of the box, then on that release we can start adding zip or exes or whatever artifacts we have.
    Simon Cozens
    @simoncozens
    Where do I find these files?
    Caleb Maclennan
    @alerque
    Trying to upload, Github just rejected it as an issue attachment, sending elsewhere.
    You can make it too with make dist.
    Simon Cozens
    @simoncozens
    Ah, OK.
    Caleb Maclennan
    @alerque
    Simon Cozens
    @simoncozens
    One of us has a bad clock. :-)
    Caleb Maclennan
    @alerque
    date ... looks right to me.
    Simon Cozens
    @simoncozens
    zip file timestamps aren't timezone aware?
    simon@upo ~/Downloads/sile-0.9.5.1.r539-g3ad2d68 $ perl -le 'print scalar localtime((stat("configure"))[9])'
    Mon Jan 13 17:56:46 2020
    simon@upo ~/Downloads/sile-0.9.5.1.r539-g3ad2d68 $ date
    Mon Jan 13 15:18:06 GMT 2020
    No, they're not. I've faked it with touch and it's building now.
    Simon Cozens
    @simoncozens
    Well that didn't work.
    make: * No rule to make target fontconfig.conf', needed by.fonts'. Stop.
    Caleb Maclennan
    @alerque
    We didn't include it in distfiles.
    Why am I not having a problem building that from the zip?
    Try 01d03b66 -> make dist ...
    Caleb Maclennan
    @alerque
    By the way you have to rerun autoconf configure.ac -f between commits in order to get the version stuff to update and hence make dist to show the right version tag. Hence the warnings about using a fully clean clone before starting in the release checklist.
    The autoconf stuff is really uncomfortable with being meddled with after it runs once. It tries way too hard to cache its status.
    Simon Cozens
    @simoncozens
    Building from dist tarball now.
    Caleb Maclennan
    @alerque
    I just did that, now building from Arch Linux package spec.
    Simon Cozens
    @simoncozens
    Looking good so far in terms of build issues - it's running the tests.
    And tests pass.