| Age | Commit message (Collapse) | Author | Lines |
|---|
|
Some host Make systems may not allow automatic passing of the number of
threads to sub-Makes. So while building the basic dependencies, we'll need
to explicity add the `-j' option to the Make files that can benefit most
from it: those that are dependencies of many others (Tar & Make), or are
the last to build (Coreutils).
|
|
The comment above the downloader section of the configure script was not up
to date with how the pipeline uses a downloader during configuration and
building now. So it was updated.
|
|
Until now we had constrained the configuration step to one thread to easily
see failures on other systems. But with most tests passing successfully
now, we are using the total number of available threads.
|
|
The build systems of Libgit2 and WCSLIB on Mac OS does not account for
installation in non-standard addresses: `Libgit2' keeps the absolute
address of its build directory (not the installation directory) and WCSLIB
doesn't write any absolute address at all (so the system uses the first one
it finds).
To address these issues, we are now using Mac OS's `install_name_tool'
program to fix the absolute path within the installed shared library.
Since the version of the library is actually present in its shared library
name, in `dependency-versions.mk' we have also separated these two
libraries so later when their version is changed, we are careful in
correcting the shared library name also.
|
|
OpenSSL can't automatically detect the architecture of Mac OS systems, so
as it suggests on its Wiki, it needs some help for doing that. With this
commit, we are checking the build on Mac OS with the presence of `otool'
(Mac OS's linker). If it's there, we'll add the OpenSSL configuration
options suggested by OpenSSL's Wiki.
|
|
Until now, we weren't including the `rpath' linking options to the basic
dependencies. They are now added. Also, when the download of an input file
fails for any reason, an empty file won't be replaced there any more.
|
|
To enable easy downloading of HTTPS links with Wget (this pipeline's defaut
downloader), we need a set of trusted CA certificates. Until the time that
we can generate one ourselves, one generic set of trusted CA certificates
is now downloaded like a tarball and placed in the OpenSSL configuration
directory.
With these CA certificates, within the pipeline we can now safely use the
pipeline's own installed Wget.
|
|
Some high-level programs like Wget and cURL need to be built in shared mode
because they also include dynamic loading of libraries. Therefore, if we
only build the lower-level libraries in static mode, our own build will be
ignored and they will go and find the system's shared libraries to link
with. Because of this, for now, we have manually set the `static_build'
variable in the configure script to `no'.
Also, if the downloader fails, we'll delete the output (an empty file in
the case of Wget) because it interefers with a target definition.
|
|
To help in debugging, we are only running the Makefiles within the
configure script on one thread.
|
|
The TeX Live installer needs Wget to operate smoothly, especially on recent
Mac OS systems that don't have Wget pre-installed. Also, it would be good
for the pipeline to have its own downloader. So with this commit, the
pipeline also installs Wget and OpenSSL which is a dependency.
Many other small changes/fixes were done in this process.
|
|
Thanks to the check by Cristina Martínez, some corrections were made when
we attempt to download and install TeXLive. Further checks and corrections
will be in due time.
|
|
The main reason I wasn't using cURL as a downloading tool was that I wasn't
familar with how to ask it to follow a re-direct. But I just found out that
its with the `-L' configure time option. So it is now added as a downloader
tool to the pipeline.
|
|
On the Libgit2 webpage, it has recommended to build it statically on Mac
systems. By default we are doing this on Linux systems, but the `-static'
flag failed on Mac. But apparently CMake might be able to deal with the
issue in a different way.
|
|
Thanks to a test build on Raul Infante Sainz's Mac OS computer, we were
able to address some issues and will be trying them after this commit:
a) The LLVM linker on that computer didn't recognize `-rpath-link'! So at
configure time we now check for it and only include it when the linker
recognizes it.
b) CMake corrections: 1) `CMAKE_LIBRARY_PATH' is now defined so CMake can
look in our custom directory to find the necessary libraries. 2) To
build and install the CMake built programs, we now simply use `make'
and `make install'.
c) To avoid particular linking problems with WCSLIB (which has special
problems compared to other libraries), we are now deleting the shared
library version (both on GNU and Mac systems).
|
|
GNU Binutils (which provides the GNU Linker) is not ported to Mac OS
systems. GCC also takes a very long time to build, and if we are to still
have linking problems with LLVM's linker, it would be better to just ignore
GCC also and use the system's C compiler and linker together.
So for the time being, GCC isn't a main target of the basic dependencies
and won't be installed. But we have kept the rules that were checked on a
GNU/Linux operating system.
|
|
The pipeline now installs GCC and all its necessary prerequisites.
|
|
We had forgot to add the rule to build the lock file directory for
downloading data. This has been corrected.
|
|
To avoid redundant steps in the the top-level Makefile and make it simpler
and easier to follow, we now define the base names of all the Makefiles in
the `makesrc' variable of the top-level Makefile. `makesrc' is then used to
define the Makefiles to include and the necessary TeX macros at the same
time. This is much more clear and obvious than the previous case were we
had to list the Makefiles and TeX macro files separately in the top level
Makefile.
|
|
While preparing the input dataset downloading scripts, we wanted to have
two input datasets, but it was too much, so it was ultimately removed. But
I had forgot to remove the information about this file in the `./configure'
script. So it is removed now.
|
|
In most analysis situations (except for simulations), an input dataset is
necessary, but that part of the pipeline was just left out and a general
`SURVEY' variable was set and never used. So with this commit, we actually
use a sample FITS file from the FITS standard webpage, show it (as well as
its histogram) and do some basic calculations on it.
This preparation of the input datasets is done in a generic way to enable
easy addition of more datasets if necessary.
|
|
Until now, in the instructions, we were suggesting to run
`./.local/bin/make', but the `./' part is extra: this is already a
directory and so the shell will be able to find it. So to make things more
clear and easy to read/write, we removed the `./' part from the calls to
our custom Make installation.
|
|
By default, Wget uses the same time stamp for the downloaded file as the
server. As a result, if a dependency was previously built before the
tarball was uploaded to the server, the pipeline wouldn't recognize that
its new and wouldn't re-build it. With this commit, we are adding the
`--no-use-server-timestamps' to the call to Wget to fix this problem.
I also noticed that (probably as a bug left over from the time we were
testing the configure script), we were manually setting the `userread'
variable to `n' independent of what the user provided. This has also been
fixed.
|
|
When you point to this project, the `README.md' file is the default file
that opens on GitLab and other online git repositories. Since a
reproduction pipeline project is different from the actual pipeline, its
best for the default text that opens to describe the paper, not the
pipeline. The old `README.md' is also kept, but its now called
`REAME-pipeline.md'.
|
|
When there is a problem in creating the final TeX Live installation, the
previous version of the pipeline would not understand and just finish! We
would later have problems in building the paper.
So the following series of steps were taken: to keep the recipes in a
shorter and easier to understand way, the steps to install TeX Live are now
one rule (that produce `.local/bin/texlive-ready-tlmgr' when its
successful), and the steps to install the necessary packages are in another
rule (that produce `.local/bin/texlive-ready' when its successful).
When control comes back inside configure, if `.local/bin/texlive-ready'
isn't there (something failed during the TeX Live installation, or building
packages), then the whole TeX Live installation directory
(`.local/texlive') will be deleted along with the two output files. This
will help ensure that future steps can check the availablility of a working
TeX Live in the pipeline.
|
|
After installing the basic dependencies, we have access to the internally
built `nproc' program (by Coreutils) to find the number of threads we can
use for building the high-level batch of dependencies. Unfortunately, I had
added an extra `./' before `$instdir' when calling `nproc' so the script
failed. This problem is fixed now.
|
|
The system's libraries are no longer used in building the higher-level
dependencies. Also, thanks to Raul Infante Sainz, we found out that Bash's
build script was still removing the extra directory information (not
good!).
|
|
When the C compiler is not GNU GCC, linking with GNU Binutils is going to
cause problems. So until the time that we can include GCC into this
pipeline, its best to avoid Binutils also.
Also, for building CMake, we were relying on an installed CMake, but now,
we are using its own `./bootstrap' script, so it can be built even if the
host system doesn't have CMake.
Also, for TeX Live, we are now setting a custom file as main target to
avoid complications with symbolic links as targets in Make.
Finally, when the user says they don't want to re-write an existing
configuration file, no extra notices will be printed and the configure
script will immediately start building programs.
|
|
To help in debugging on other systems, the building of dependencies is only
done on a single thread.
|
|
Since the final product of the pipeline is a LaTeX-created PDF file, it was
necessary to also have LaTeX within the pipeline. With this commit, TeX
Live is also built as part of the configuration and all the necessary
packages to build the PDF are also installed and mentioned in the paper
along with their versions.
|
|
TeX Live is now also downloaded and built by the reproduction
pipeline. Currently on the basic (TeX and LaTeX) source is built but no
extra packages, so the PDF building will fail. We'll add them in the next
commit.
|
|
In the previous commit, for testing the static build, I had added a
`-ljunk' option to the compiler. But I had forgot to remove it! It is
removed now.
|
|
The default Mac compiler has problems building static libraries. Since we
are not yet building the GNU C Compiler as part of the pipeline, we'll have
to rely on the host system's compiler. Therefore, a check is now added a
the start of the configure script that will build a minimal program with
the `-static' flag and if it fails, it will print a warning. Afterwards,
none of the dependencies will be built with the `-static' flag.
|
|
To have better control over the build, GNU Binutils, Bzip2, GNU Gzip, and
XZ Utils have also been added to the pipeline. Some other minor cleanups
and fixes were also implemented throughout the process.
|
|
Until now, when a package was to be built statically, we were adding the
`--static' option to `CFLAGS'. This was the wrong place to put it! It
should be in the linking step (thus `LDFLAGS'). Also, based on Bash's
configure script, we are now using the more generic form of `-static'
(single dash, not double dash).
On the other hand, the `--disable-shared' option isn't available in many of
the packages and it is highly redundant with the `-static' option, so it
has been removed to avoid an extra warning in such packages.
|
|
To ensure the easy unpacking and building of the programs, Lzip and Tar are
now also build during the initial setup phase.
Some minor corrections were also applied to make things cleaner and
smoother.
|
|
Until now, we used semicolons in Make's Call function definitions to build
the programs with GNU build system or CMake. Therefore, if any step of the
process failed, the rest would be ignorant to it and pass. Now, we use `&&'
to separate the different processing steps. In this way, we can be sure
that if any of them fails (during configuration, or building for example),
the pipeline will also stop and not continue to the next command (in the
same recipe).
Since the two Make Call functions were identical in the two
`dependencies-basic.mk' and `dependencies.mk', they are now in one file to
be imported in both.
This bug was found by Raul Infante Sainz.
|
|
After a test by Raúl Infante Sainz, we found out that the configure script
and the Make script for Bash and Make are making too many assumptions on
more recent versions of both. As a result, it couldn't be built.
Therefore, the `configure' script was modified to not use more recent tools
like `readlink' (to find the absolute address of a relative one). It was
also re-organized to not have to read the configuration parameters from a
text file. The parameters are directly read from the command-line and are
written into the proper file afterwards. This removes the need to opening a
text editor by the user (which also caused problems on Raúl's system).
To fix the Make version issue, the building of Bash and Make are now done
in a new Makefile (`reproduce/src/make/dependencies-basic.mk'). This file
doesn't make many of the assumptions that were made in
`dependencies.mk'. So it should hopefully work on any version of Make.
To help in debugging, for now, the Makefile of configure, are asked to work
on one thread (the `-j' option is commented in the `configure'). But after
checks, we'll fix this.
|
|
During the configuration step several new programs that were necessary for
a more complete controlled environment are now also downloaded and built
statically.
|
|
To enable easy/proper reproduction of results, all the high-level
dependencies are now built within the pipeline and installed in a fixed
directory that is added to the PATH of the Makefile. This includes GNU Bash
and GNU Make, which are then used to run the pipeline.
The `./configure' script will first build Bash and Make within itself, then
it will build
All the dependencies are also built to be static. So after they are built,
changing of the system's low-level libraries (like C library) won't change
the tarballs.
Currently the C library and C compiler aren't built within the pipeline,
but we'll hopefully add them to the build process also.
With this change, we now have full control of the shell and Make that will
be used in the pipeline, so we can safely remove some of the generalities
we had before.
|
|
We had previously started the `configure' script with `/bin/bash'. But this
script is meant to check for Bash inside of it. So to be run-able (on a
system which may not have Bash), the `configure' script has to be run by
`/bin/sh'.
|
|
Different implementations of AWK may use different random number
generators, so even setting the seed will not ensure a reproducible
result. Because of this, the random plot may be different when the
pipeline runs on different systems and this can confuse early users
(its contrary to the exact reproducibility that is the whole purpose
of this pipeline).
The plot is just a simple X^2 plot, showing the squared value of the X
axis on the Y axis. It is very simple, but atleast it will be
identical on all systems. Also, there may be too many complicated
things in the pipeline already for an early user, and its just a
demonstration, so the easier/simpler, the better.
|
|
While we had set the default Makefile SHELL to be bash, we weren't
actually checking if `bash' is available on the system. With this
commit, it is also checked at configure time.
|
|
Until now, the copyright statement was left empty for the users of the
pipeline to fill. However, the files have already been created and have an
author (or contributing authors) before the user starts using the
pipeline. So the original authors of the files are added along with the
year. The user can add their own name to the existing files under the
"Contributing author" when they start and they will be the "Original
author" of the new files they create.
Several changes were also made to the TeX management:
- LaTeX is run within a `reproduce/build/tex/build' directory now. Not in
the top reproduction pipeline directory. This helps keep all the
auxiliary TeX files and directories in that directory and keep the top
reproduction pipeline directory clean. After the final PDF is built, a
copy is put in the top reproduction pipeline directory for easy viewing.
- The PGFPlots preamble was also made more useful, allowing the name of
the `.tex' file to also be the name of the final plot that is
produced. This is a GREAT feature, because without it, the TiKZ
externalization would be based on order of the plots within the
paper. But now, order is irrelevant and we can even delete the TiKZ
files within the processing workhorse-Makefiles so the plots are
definitly rebuilt on the next run.
- The paper is now in a two-column format to be more similar to published
papers.
A tip on debugging Make was added to `README.md'.
|
|
The mandatory and optional (for example downloader) dependencies are now
checked at configure time so users can know what they may be missing before
the processing starts. Since its recommended to be run in parallel, it can
be hard to find what you are missing after running the pipeline. As part of
these checks, the program to use for downloading is now also set at
configure time, it is only used as a pre-defined (in `LOCAL.mk') variable
during Make's processing.
A small title was also added to discus the pipeline architecture that will
be filled in the next commit.
|
|
Some editors put a copy of their input file into another file ending with
`~' (for backup). So now, the `./configure' script also cleans this file
along with the temporary file.
|
|
While trying the pipeline on a remote server (which runs on Debian), the
configure script had an `Syntax error: "(" unexpected' error. This is
caused by the fact that in the Debian world (and its derivate OSs), the
default shell is not Bash but Dash which has much fewer features for fast
loading. It was thus necessary to start the configure script explicity with
the `/bin/bash' shebang.
|
|
As described in the commens above `MINMAPSIZE' of `LOCAL.mk.in', the amount
of memory to map to HDD/SSD or keep in RAM is a local issue and not
relevant to the pipeline's results. So it is now defined in a
`gnuastro-local.conf' file.
To keep the Makefiles clean, this file is created by the `./configure'
script. To do this cleanly, the `./configure' script was also almost fully
re-written with better functionality now.
|
|
The previous change where we had set the building of the PDF as a local
(and thus not version controlled) setting was not good, because different
commits might be made without the high-level preparations for the final PDF
(especially during the initial/testing phases of a research). Therefore, if
the runner of the pipeline is ignorant to this, they may hit some errors in
LaTeX which can be frustrating.
To have a clean reproduction, it is thus necessary to have the choice of
pdf-building under version control along with the rest of the pipeline.
|
|
The choice of whether or not to make a PDF is now also a local system
issue, not a general pipeline issue. So it has been put in the new
`LOCAL.mk.in' file which replaces the old `DIRECTORIES.mk.in'. All local
settings (things that when changed should not be version-controlled) should
be defined in this file.
A sanity check was added to find if `./configure' has been run before
`make' or not (using the `LOCAL.mk' file which is an output of the
configuration step). If `LOCAL.mk' doesn't exist, an error will be printed
informing the user that `./configure' needs to be run first.
The configure script also provides more clear and hopefully better
information on its purpose and what must be done.
Since `make clean', it is executed even when `./configure' hasn't been run,
it will only delete the build directory and its contents when local
configuration has been done.
A `distclean' target was also added which will first "clean" the pipeline,
then delete the `LOCAL.mk.in' file.
To allow rules like `make' to be run even if `BDIR' isn't defined
(`./configure' hasn't been run yet), a fake `BDIR' is defined in such
cases.
|
|
Let's start working on this pipeline independently with this first
commit. It is based on my previous experiences, but I had never made a
skeleton of a pipeline before, it was always within a working analysis.
But now that the pipeline has a separate repository for its self, we will
be able to work on it and use it as a base for future work and modify it to
make it even better. Hopefully in time (and with the help of others), it
will grow and become much more robust and useful.
|
