SQLAlchemy is currently at version 0.7.6: http://docs.sqlalchemy.org/en/latest/ though the PDF build (http://media.readthedocs.org/pdf/sqlalchemy/latest/sqlalchemy.pdf), while it seems to have the correct date of "March 14, 2012" inside the file and additionally has up-to-date content, still displays "0.7.5" on the title page, confusing and misleading our users. I'd gather this is due to some of the state from the previous versions' PDF build still hanging around. Is there an ETA on the "really, really, blow everything away and start again" button ? Why isn't at least the naming convention used by Sphinx itself used ? That is, sqlalchemy_0_7_6.pdf ?
Hey Mike, Yea, we basically archive everything that ever gets built on the media servers. There isn't really a great automated way of knowing when something might need to be removed or stored forever. Having this in the UI might be a good idea (eg. more fine-grained editing about what stays around). As for the file names, there is no particular reason that aren't versioned. It seems dumb in hindsight, and should likely be fixed. I will look into at least copying the files over with version numbers in the files. I will probably need to keep the old <slug>.pdf files around for people who are linking against them, or I can likely just symlink them to the newer, better file names. Cheers, Eric On Tue, Mar 20, 2012 at 8:51 AM, Michael Bayer <mike_mp@zzzcomputing.com>wrote: > SQLAlchemy is currently at version 0.7.6: > > http://docs.sqlalchemy.org/en/latest/ > > though the PDF build ( > http://media.readthedocs.org/pdf/sqlalchemy/latest/sqlalchemy.pdf), while > it seems to have the correct date of "March 14, 2012" inside the file and > additionally has up-to-date content, still displays "0.7.5" on the title > page, confusing and misleading our users. > > I'd gather this is due to some of the state from the previous versions' > PDF build still hanging around. Is there an ETA on the "really, really, > blow everything away and start again" button ? > > Why isn't at least the naming convention used by Sphinx itself used ? > That is, sqlalchemy_0_7_6.pdf ? > > > > > -- Eric Holscher Engineer at Urban Airship in Portland, Or http://ericholscher.com
It looks like you now have the "wipe out the build" command: http://read-the-docs.readthedocs.org/en/latest/builds.html#deleting-a-stale-or-broken-build-environment http://readthedocs.org/wipe/sqlalchemy/latest/ I've run that and am rebuilding to see if the PDF gets rebuilt completely. On Mar 26, 2012, at 4:17 PM, Eric Holscher wrote: > Hey Mike, > > Yea, we basically archive everything that ever gets built on the media servers. There isn't really a great automated way of knowing when something might need to be removed or stored forever. Having this in the UI might be a good idea (eg. more fine-grained editing about what stays around). > > As for the file names, there is no particular reason that aren't versioned. It seems dumb in hindsight, and should likely be fixed. I will look into at least copying the files over with version numbers in the files. I will probably need to keep the old <slug>.pdf files around for people who are linking against them, or I can likely just symlink them to the newer, better file names. > > Cheers, > Eric > > On Tue, Mar 20, 2012 at 8:51 AM, Michael Bayer <mike_mp@zzzcomputing.com> wrote: > SQLAlchemy is currently at version 0.7.6: > > http://docs.sqlalchemy.org/en/latest/ > > though the PDF build (http://media.readthedocs.org/pdf/sqlalchemy/latest/sqlalchemy.pdf), while it seems to have the correct date of "March 14, 2012" inside the file and additionally has up-to-date content, still displays "0.7.5" on the title page, confusing and misleading our users. > > I'd gather this is due to some of the state from the previous versions' PDF build still hanging around. Is there an ETA on the "really, really, blow everything away and start again" button ? > > Why isn't at least the naming convention used by Sphinx itself used ? That is, sqlalchemy_0_7_6.pdf ? > > > > > > > > -- > Eric Holscher > Engineer at Urban Airship in Portland, Or > http://ericholscher.com
OK, it rebuilt, but not correctly. pdflatex needs to be run multiple times in order to generate the table of contents. The Makefile that comes with Sphinx does this: %.pdf: %.tex -pdflatex $(LATEXOPTS) '$<' -pdflatex $(LATEXOPTS) '$<' -pdflatex $(LATEXOPTS) '$<' -makeindex -s python.ist '$(basename $<).idx' -pdflatex $(LATEXOPTS) '$<' -pdflatex $(LATEXOPTS) '$<' In SQLAlchemy's case, we also pass the flag -interaction=nonstopmode so that the build runs all the way through. It seems, though I can't tell for sure, like readthedocs doesn't use the Makefile, and also annoyingly doesn't show me the commands it run within the logs to even check what it's doing. Assuming that's the case, there's no way to affect how RTD runs the build commands and there should be. On Mar 28, 2012, at 10:59 AM, Michael Bayer wrote: > It looks like you now have the "wipe out the build" command: > > http://read-the-docs.readthedocs.org/en/latest/builds.html#deleting-a-stale-or-broken-build-environment > http://readthedocs.org/wipe/sqlalchemy/latest/ > > I've run that and am rebuilding to see if the PDF gets rebuilt completely. > > > > On Mar 26, 2012, at 4:17 PM, Eric Holscher wrote: > >> Hey Mike, >> >> Yea, we basically archive everything that ever gets built on the media servers. There isn't really a great automated way of knowing when s omething might need to be removed or stored forever. Having this in the UI might be a good idea (eg. more fine-grained editing about what stays around). >> >> As for the file names, there is no particular reason that aren't versioned. It seems dumb in hindsight, and should likely be fixed. I will look into at least copying the files over with version numbers in the files. I will probably need to keep the old <slug>.pdf files around for people who are linking against them, or I can likely just symlink them to the newer, better file names. >> >> Cheers, >> Eric >> >> On Tue, Mar 20, 2012 at 8:51 AM, Michael Bayer <mike_mp@zzzcomputing.com> wrote: >> SQLAlchemy is currently at version 0.7.6: >> >> http://docs.sqlalchemy.org/en/latest/ >> >> though the PDF build (http://media.readthedocs.org/pdf/sqlalchemy/latest/sqlalchemy.pdf), while it seems to have the correct date of "March 14, 2012" inside the file and additionally has up-to-date content, still displays "0.7.5" on the title page, confusing and misleading our users. >> >> I'd gather this is due to some of the state from the previous versions' PDF build still hanging around. Is there an ETA on the "really, really, blow everything away and start again" button ? >> >> Why isn't at least the naming convention used by Sphinx itself used ? That is, sqlalchemy_0_7_6.pdf ? >> >> >> >> >> >> >> >> -- >> Eric Holscher >> Engineer at Urban Airship in Portland, Or >> http://ericholscher.com >
right so on the second forced build without a clean of the environment, the pdf builds correctly. Can I please have the "-interaction=nonstopmode" flag hardcoded into the options used with LATEXOPTS for now and ensure that pdflatex is called multiple times? I still need to do a manual wipe on each new version number but at least this would have the pdf build "passing". On Mar 28, 2012, at 11:14 AM, Michael Bayer wrote: > OK, it rebuilt, but not correctly. pdflatex needs to be run multiple times in order to generate the table of contents. The Makefile that comes with Sphinx does this: > > %.pdf: %.tex > -pdflatex $(LATEXOPTS) '$<' > -pdflatex $(LATEXOPTS) '$<' > -pdflatex $(LATEXOPTS) '$<' > -makeindex -s python.ist '$(basename $<).idx' > -pdflatex $(LATEXOPTS) '$<' > -pdflatex $(LATEXOPTS) '$<' > > In SQLAlchemy's case, we also pass the flag -interaction=nonstopmode so that the build runs all the way through. > > It seems, though I can't tell for sure, like readthedocs doesn't use the Makefile, and also annoyingly doesn't show me the commands it run within the logs to even check what it's doing. Assuming that's the case, there's no way to affect how RTD runs the build commands and there should be. > > > > > On Mar 28, 2012, at 10:59 AM, Michael Bayer wrote: > >> It looks like you now have the "wipe out the build" command: >> >> http://read-the-docs.readthedocs.org/en/latest/builds.html#delet ing-a-stale-or-broken-build-environment >> http://readthedocs.org/wipe/sqlalchemy/latest/ >> >> I've run that and am rebuilding to see if the PDF gets rebuilt completely. >> >> >> >> On Mar 26, 2012, at 4:17 PM, Eric Holscher wrote: >> >>> Hey Mike, >>> >>> Yea, we basically archive everything that ever gets built on the media servers. There isn't really a great automated way of knowing when s omething might need to be removed or stored forever. Having this in the UI might be a good idea (eg. more fine-grained editing about what stays around). >>> >>> As for the file names, there is no particular reason that aren't versioned. It seems dumb in hindsight, and should likely be fixed. I will look into at least copying the files over with version numbers in the files. I will probably need to keep the old <slug>.pdf files around for people who are linking against them, or I can likely just symlink them to the newer, better file names. >>> >>> Cheers, >>> Eric >>> >>> On Tue, Mar 20, 2012 at 8:51 AM, Michael Bayer <mike_mp@zzzcomputing.com> wrote: >>> SQLAlchemy is currently at version 0.7.6: >>> >>> http://docs.sqlalchemy.org/en/latest/ >>> >>> though the PDF build (http://media.readthedocs.org/pdf/sqlalchemy/latest/sqlalchemy.pdf), while it seems to have the correct date of "March 14, 2012" inside the file and additionally has up-to-date content, still displays "0.7.5" on the title page, confusing and misleading our users. >>> >>> I'd gather this is due to some of the state from the previous versions' PDF build still hanging around. Is there an ETA on the "really, really, blow everything away and start again" button ? >>> >>> Why isn't at least the naming convention used by Sphinx itself used ? That is, sqlalchemy_0_7_6.pdf ? >>> >>> >>> >>> >>> >>> >>> >>> -- >>> Eric Holscher >>> Engineer at Urban Airship in Portland, Or >>> http://ericholscher.com >> >