| Age | Commit message (Collapse) | Author | Lines |
|---|
|
One major point was that following Konrad's suggestion the issue of not
being familiar with the Lisp/Scheme framework of GWL is now removed. We
actually mention the main problem we have had with Guix, but also highlight
that their solution was one of the main inspirations for this work.
|
|
Until this commit, when the user have previous TeX tarball already
present, the project crashed when trying to re-configure, if there was a
newer version of TeX. This is because TeX are updated yearly. With this
commit, this bug has been fixed. Now, during the installation of TeX, it
checks if this problem happens. If this is the case, then it moves the
old tarball, download the new one and install it. If not, it will just
install the already present tarball or crash because of any other
reason. This probem was recurrent, and each time TeX was updated, the
previous tarball had to be removed manually. But now, with this commit,
it is done automatically. The detection and fix of this bug has been
possible with the help of Mohammad Akhlaghi, thanks!
|
|
Until this commit, there was only a small description of me. With this
commit, I have added a small paragraph with my biography. I know we are
very restricted because of the word limit so I tried to be very short!
|
|
With this commit, I have corrected several minor typos.
|
|
With this commit, I did some minor changes in these Sections. Main
changes are: define the contraction `OS' from Operating System and use
only `OS' later on, and not use contractions like `isn't'
|
|
Before this commit: Roberto's bio was about 120 words. With this
commit: it is now less than 100 words. A comment about
reproducibility has been added.
|
|
Publishing a paper on reproducible research without making it easy for
readers to read the references would defeat the point. Of course we have to
make some compromises with some journals' reluctance to shift towards the
free world, but to satisfy scientific ethics, we should at least provide
clickable URLs to the references, preferably to the ArXiv version if
available [1], and also to the DOI, again, preferably to an open-access
version of the URL if available.
I was not able to fully get this done in the .bst file, so there's an
sed/tr hack done to the .bbl file in `reproduce/analysis/make/paper.mk` to
tidy up commas and spaces.
This commit also reverts some of the hacks in the Akhlaghi IAU Symposium
`tex/src/references.tex` entry, to match the improved .bst file,
`tex/src/IEEEtran_openaccess.bst`, provided here with a different name to
the original, in order to satisfy the LaTeX licence.
[1] https://cosmo.torun.pl/blog/arXiv_refs
|
|
David and Raul had both reported that because 'pdftotext' wasn't available
on their system, the project failed (even though the PDF was built!). So
with this commit, we first check if the system has 'pdftotext' and call it
only if its is available.
Some minor edits were made, building upon Boud's previous commit.
|
|
This commit provides mostly small changes. There didn't seem much point in
repeating the `lessons learned` jargon and claiming that we draw good
conclusions - insights - from our experience. Better just state what
hypotheses we have generated from the experience rather than give the
misleading impression that our hypotheses are well-established facts. In
the comments, I put a suggested translation of what the `lessons learned`
jargon means. I seem to have first heard this term in the mainstream media
a few years after the US 2003 attack on Iraq, when a US military
representative stated that the US forces had "learned lessons" after having
started a war of aggression against Iraq.
|
|
This commit changes two lines.
(1) Keeping the exact quote with the clerk while having a sentence that
makes sense in plain English cannot be done, it seems to me, without
making the sentence a bit longer. Here's one option that seems about
the best we can do, even though it still sounds a bit funny, because
it's hard to write a future conditional with the present "can". Since
it's a quote, it will probably survive the proofreaders.
(2) Software is an uncountable noun [1], so we say "software is", like
"water is"; "used software" sounds odd; I added "is itself" to
emphasise that we're especially talking about the full chain of
software for running the project. This commit modifies the "When the
..." sentence and hopefully sounds better.
[1] https://en.wiktionary.org/wiki/software#Noun
|
|
To help show the simplicity of 'top-make.mk', it was included as a
listing. I also went over some of Boud's corrections and made small
edits. In particular:
- The '\label' and '\ref' to a section were removed. I done this after
inspecting some of their recent papers and noticing that they generally
have a simple flow, without such redirections.
- In the part about the RDA adoption grant, I moved the "from the
researcher perspective" to the end. Because Austin+2017 is mainly
focused on data-center management, not the researcher's. They do touch
upon researcher solutions that can help data-base managers, but not
directly the researchers. In effect with this grant, they acknowledged
that our researcher-focused solution confirms with their criteria for
data-base management.
|
|
Possibly the least trivial edit in this commit is that the previous text
appeared to state that it's normal to find that a project prepared with
`maneage` may be ... unbuildable. Which would defeat our whole claim of
reproducibility! Obviously, `maneage` is still in a rapid development
stage and might still have significant, not-yet-detected bugs. But the
wording has to explain that this would constitute a bug in `Maneage` (in a
particular version of it), not an expected regular event. :) This commit
aims to fix that and other minor wording issues in IV.
Pdf word count 5855.
|
|
This series of commits aims to edit sections II+III, but first implements
the changes from 7bf5fcd, apart from one that conflicts in the abstract:
this commit has ``Maneage'' without `(managing+lineage)` in the abstract.
From Mohammad: this commit has been rebased after several other parallel
branches, so some things may differ from the message.
|
|
Until this commit, when the user had a previous TeXLive tarball already
present (in their software-tarball directory) compared to the CTAN server,
the project crashed in the configure phase. This was because TeXLive is
updated yearly and we don't yet install TeXLive from source (currently we
use its own package manager, but we plan to fix this in task #15267).
With this commit, we fix the problem by checking the cause of the crash
during the installation of TeX. If the crash is due to this particular
error, we ignore the old tarball and download the new one and install it
(the old one is still kept in '.build/software/tarballs', but will get a
'-OLD' in its name. This probem was recurrent, and every year that TeXLive
is updated, the previous tarball had to be removed manually! But with this
commit, this is done automatically. The detection and fix of this bug has
been possible with the help of Mohammad Akhlaghi, thanks!
|
|
One of the main reasons to building Maneage is to properly
acknowledge/attribute the authors of software in research. So we have
adopted a standard of never referring to the GNU-based operating systems
running the Linux kernel simply as "Linux", we avoid terms like "Open
Sourse" and use Free Software instead (in the same spirit).
With this commit, a few instances of the cases above have been corrected,
they had slipped through our fingers when we initially imported them into
the project. In the special case of the "Journal for Open Source Software",
we simply replaced it with its abbreviation (JOSS). This was done because
in effect we were generally using journal name abbreviations in almost all
the citations already. To avoid any inconsistancies, the names of the three
other journals that weren't abbreviated are also abbreviated.
|
|
Generally they were great, but after looking through them I thought a
hand-full of them slightly changed my original idea so I am correcting them
here. Boud, if you feel the changes aren't good, let's talk about it and
find the best way forward ;-).
They are mostly clear from a '--word-diff', just some notes on the ones
that have changed the meaning:
* On the "a clerk can do it" quotation, since its so short, I think its
better to keep its original form, otherwise a reader may thing there
were paragraphs instead of the "to" and we have changed their
intention.
* In the part where we are saying that the workflow can get "separated"
from the paper, I mostly meant to highlight that the data-centers and
journals (hosts) may diverge in decades, or one of them may go
bankrupt, or etc. Hence loosing the connection. The issue of it
evolving can in theory be addressed through version control, so I think
this is a more fundamental problem.
* In the part about free software, in the list, the original point was
the free software that are used by the project, not the project itself
(after all, the project itself falls under the "Open Science" titles
that is very fashionable these days, but my point here is to those
people who claim to do "Open Science" with closed software (like
Microsoft Excel!).
|
|
This commit makes several small changes to Section III, some of
which are quite significant in terms of meaning.
It was difficult to improve the clarity without extending
the word length. Now we're at 5901 words.
|
|
This commit implements quite a few minor changes in section II.
The aim of most is to clarify the meaning and remove ambiguity.
A few changes are that the reader will normally assume that
successive sentences in a paragraph are closely related in terms
of logical flow. It is superfluous - and considered excessive -
to put too many "Therefore"'s and "Hence"'s in (at least) modern
astronomy style. These are supposed to be used when there is a
strong chain of reasoning.
One change is done in the Introduction, because if we're going to
use "solution(s)" throughout to mean "reproducible workflow
solution(s)", then we have to clearly define this as jargon for
this particular paper. It's probably preferable to RWS - reproducible
workflow solution - or RWI - reproducible workflow implementation.
But we can't just keep saying "solution" because that has many
different meanings in a scientific context.
Pdf word count = 5880
|
|
This series of commits aims to edit sections II+III,
but first implements the changes from 7bf5fcd, apart
from one that conflicts in the abstract: this commit
has ``Maneage'' without `(managing+lineage)` in the
abstract.
|
|
This commit implements most of David's changes from c76727b, but excluding
some, such as the proposal to use 'which' in a restrictive clause in the
abstract. This is allowed, but the Fowler brothers' rule tends to followed
in science writing:
https://ell.stackexchange.com/questions/5/is-there-any-difference-between-which-and-that
A few points on the abstract:
* an immediate solution = singular
* The "immediate, fast short-term" benefits sentence sounded like it was
redundantly superfluously repetitively repeating doubled-up
information. Hopefully this edit is better.
* in the %Conclusion, "solutions" is vague, like people who say
"technology" when they're only talking about software, so this edit
reminds the reader to make the sentence more self-contained and
understandable.
|
|
With this commit, Maneage now includes instructions to build the memory
tracing tool Valgrind and the program 'patch' (to apply corrections/patches
in text files and in particular the sources of programs).
For this version of Valgrind, some patches were necessary for an interface
with OpenMPI 2.x (which is the case now). Also note that this version of
Valgrind's checks can fail with GCC 10.1.x (when using '--host-cc'), and
the failures aren't due to internal problems but due to how the tests are
designed (https://bugs.gentoo.org/707598). So currently if any of
Valgrind's checks fail, Maneage still assumes that Valgrind was built and
installed successfully.
While testing on macOS, we noticed that it needs the macOS-specific 'mig'
program which we can't build in Maneage. DESCRIPTION: The mig command
invokes the Mach Interface Generator to generate Remote Procedure Call
(RPC) code for client-server style Mach IPC from specification files. So a
symbolic link to the system's 'mig' is now added to the project's programs
on macOS systems.
This commit's build of Patch and Valgrind has been tested on two GNU/Linux
distributions (Debian and ArchLinux) as well as macOS.
Work on this commit started by Boud Roukema, but also involved tests and
corrections by Mohammad Akhlaghi and Raul Infante-Sainz.
|
|
David reported this problem, it happened right after importing IEEEtran,
but for some reason, it didn't happen for me.
|
|
When entering the name of the "listings" package, I had forgot to add the
final 's', so it wasn't being installed on a clean system! I didn't have a
problem until now, because it remained from previous builds.
|
|
After a look at the PDFs of the linked papers of the previous commit and a
few 2020 papers, we noticed that the biography format of the webpage and
PDFs are different! So it is now back in its old way (which is how
biographies are presented in the PDF).
A few other minor edits were made in the text.
|
|
It appears from looking at
https://ieeexplore.ieee.org/document/5725236/authors#authors
https://ieeexplore.ieee.org/document/7878935/authors#authors
that the affiliations section needs to start with a one-phrase
definition of the author's main affiliation. In 5725236,
the typesetters/proofreaders swapped van der Walt and Colbert,
so don't be confused by that. It shows that nobody proofread
properly.
With this commit, each author's institute (single hierarchical
level) is written as the first paragraph of the author's
affiliation section. Since 5725236 allows a very-well-known
acronym, I'm guessing that IAC can be defined for Mohammad
and then re-used for the others.
I've added a brief CV for me. If necessary, we could compress
my main research together as "observational cosmology", but let's
see how we go in the word count.
I have not (yet) worked through the main text.
There is also one minor language fix - `Because is complete` was
incomplete.
Pdf word count: 5873
|
|
After one day not looking at the first draft of this new version (commit
7b008dfbb9b2), I went through the text and done some general edits to make
its presentation and logic smoother.
|
|
Before this commit: several typos were present along the text. With this
commit several typos have been corrected (types listed below) and my bio
has been added.
a) double words
b) general typos
c) comas after adverbs at the beginning of a sentence
d) contractions are removed, e.g., don't vs do not
e) three sentences in parenthesis have been removed since I think they
were out of context or unnecessary
f) etc
|
|
In order to correspond to the updated datalineage plot, the name of the
plotted columns was changed to 'columns.txt', but I had forgot to update it
in the LaTeX source and since the old file still remained I hadn't
noticed. This was found by Boud and corrected.
|
|
Until now, two of the software BibTeX sources (Matplolib and Sympy) had an
"abstract" entry that was long, not similar to the rest, and not relevant
in this context, so they are removed with this commit.
|
|
In time, some of the copyright license description had been mistakenly
shortened to two paragraphs instead of the original three that is
recommended in the GPL. With this commit, they are corrected to be exactly
in the same three paragraph format suggested by GPL.
The following files also didn't have a copyright notice, so one was added
for them:
reproduce/software/make/README.md
reproduce/software/bibtex/healpix.tex
reproduce/analysis/config/delete-me-num.conf
reproduce/analysis/config/verify-outputs.conf
|
|
Following the fact that the DSJ editor decided that this paper doesn't fit
into their scope, we decided to submit it to IEEE's Computing in Science
and Engineering (CiSE). So with this commit the text was re-written to fit
into their style and word-count limitations.
|
|
Until this commit, when the version of Gnuastro doesn't match with the
version that the project was designed to use, the warning message saying
how to run the configure step was not showing the option `-e'. This
situation is normal when updating the version of Gnuastro to the most
recent one (with the project already configured). However, the use of this
option is more convenient than giving the top-build directory, etc, every
time. With this commit, the warning message has been changed in order show
also the option `-e' in the re-configure of the project.
|
|
Until this commit, Scamp was installed with the option
`--enable-plplot=yes' (the default). However, Maneage does not have PLplot
included. As it is possible to install Scamp without PLplot (in that case
it won't generate plots), with this commit this option has been set to
`no'. As a consequence, Scamp will be installed even if the host system
does not have PLplot without crashing (but it won't make any plot).
|
|
Until now Maneage used the host's GNU Gettext if it was present. Gettext is
a relatively low-level software that enables programs to print messages in
different languages based on the host environment. Even though it has not
direct effect on the running of the software for Maneage and the lanugage
environment in Maneage is pre-determined, it is necessary to have it
because if the basic programs see it in the host they will link with it and
will have problems if/when the host's Gettext is updated.
With this commit (which is actually a squashed rebase of 9 commits by Raul
and Mohammad), Gettext and its two extra dependencies (libxml2 and
libunistring) are now installed within Maneage as a basic software and
built before GNU Bash. As a result, all programs built afterwards will
successfully link with our own internal version of Gettext and
libraries. To get this working, some of the basic software dependencies had
to updated and re-ordered and it has been tested in both GNU/Linux and
macoS.
Some other minor issues that are fixed with this commit
- Until this commit, when TeX was not installed, the warning message
saying how to run the configure step in order to re-configure the
project was not showing the option `-e'. However, the use of this option
is more convenient than entering the top-build directory and etc every
time. So with this commit, the warning message has been changed in order
use the option `-e' in the re-configure of the project.
- Until now, on macOS systems, Bash was not linking with our internally
built `libncurses'. With this commit, this has been fixed by setting
`--withcurses=yes' for Bash's configure script.
|
|
Until now there we had manually inserted a `\' before the `_' of sip_tpv
program. However, we also recently added a step in the configure script to
add a `\' before every `_' when writing the final LaTeX macro. This was
because some C compilers (when the host's is used) have an `_' in their
version that we had no control over.
With this commit, the `\' is removed from `sip_tpv' in its build-rule and
we let the backslash be inserted automatically.
|
|
Until now, when you changed the version of a software in an already-built
system, its tarball would be downloaded, but it wouldn't actually
build. The only way would be to force the build by deleting the main target
of that file (under `.local/version-info/TYPE/PROGRAM'). This was because
the tarballs were an order-only prerequisite which was implemented some
time ago based on some theoretical argument that if the tarball dates
changes, the software should not be rebuilt (because we check the
checksum).
However, the problems this causes are more than those it solves: Users may
forget to delete the main target of the program and mistakenly think that
they are using the new version. The fact that all the numbers going into
the paper also contain this number further hides this.
With this commit, tarballs are no longer order-only and any time a version
of a software is updated, it will be automatically built and not cause
confusion and manual intervention by the users. As a result of this change,
I also had to correct the way we find the tarball from the list of
prerequisites.
|
|
The paper is no longer using LuaLaTeX, but raw LaTeX (that saves a DVI), it
is so much faster! Initially I had used LuaLaTeX to use special fonts to
resemble the CODATA Data Science Journal, but all that overhead is no
longer necessary. Therefore I also removed the MANY extra LaTeX packages we
were importing. The paper builds and is able to construct one of its images
(the git-branching figure) with only 7 packages beyond the minimal
TeX/LaTeX installation. Also in terms of processing it is so much faster.
The text is just temporary now, and mainly just a place holder. With the
next commit, I'll fill it with proper text.
|
|
As explained in the new `README-hacking.md', this files greatly helps in
avoiding un-necessary conflicts.
|
|
A few small conflicts showed up here and there. They are fixed with this
merge.
|
|
Until this commit, the configure step would fail with an error when
compiling libgit2 on a test system. The origin of this bug, on the OS that
was tested, appears to be that in OpenSSL Version 1.1.1a, openssl/ec.h
fails to include openssl/openconf.h. The bug is described in more detail at
https://savannah.nongnu.org/bugs/index.php?58263
With this commit, this is fixed by manually inserting a necessary
components. In particular, `sed` is used to insert a preprocessor
instruction into `openssl/openconf.h`, defining `DEPRECATED_1_2_0(f)`, for
an arbitrary section of code `f`, to include that code rather than exclude
it or warn about it.
This commit is valid provided that openssl remains at a version earlier
than 1.2.0. Starting at version 1.2.0, deprecation warnings should be run
normally. We have thus moved the version of OpenSSL in `versions.conf' to
the section for programs that need to be manually checked for version
updates with a note to remind the user when reaching that version.
Other packages that use OpenSSL may benefit from this commit, not just
libgit2.
|
|
The difference between `that` and `which` is not strictly
required, but it helps clarify the difference in meaning,
which is important in science and software :).
This is best shown by an example:
* Maneage provides reproducibility, which is a good thing.
The sentence would make sense if we drop `, which is a good
thing.` The last part of the sentence is a comment rather than a
necessary part of the sentence.
* Maneage provides a quality of reproducibility that is missing
from other implementations.
The sentence would not quite make sense if we drop `that is ...`,
since we would not know what sort of quality is provided. The
fact that the quality is missing is key to the intended meaning
of the sentence.
|
|
It is also slightly shorter with this commit, without loosing anything
substantial.
|
|
No need to invent a new word (archive-able) when an existing one
(archivable) does the job.
One issue that we have not included and which perhaps we could discuss in
the paper (space permitting), is that this tool could bypass the use of
blockchains in this context.
|
|
As discussed by Boud in the previous commit, this is an important feature
that was lost in the new abstract. So I added it as a criteria.
|
|
Most are minor English tidying, e.g.
* spelling: achieving
* archivable - https://en.wiktionary.org/wiki/archivable
* `i.e.` does not look good in an abstract;
* `when` didn't sound quite right;
Comment: we no longer state one of the most interesting aspects
of Maneage - producing the draft paper that is submittable for
peer review in a way that makes it natural for the authors to
achieve automatic consistency between the calculations/analysis
and the values in the paper. But this is hard to describe in a
compact way without disrupting the overall argument of the
abstract, so it's a bit of a pity, but people will learn about it
anyway from the body of the article (or from trying out the
package!) `Peer-review verification` does not directly state
producing a pdf.
Related to this absence of talking about reproducing the *paper*,
not just the calculations, I suggest dropping `, with snapshot
\projectversion` from the abstract initially sent to the journal
(they can't stop us updating it afterwards), because without the
context of explaining that the paper itself is produced from the
package, it's not clear what the snapshot means - a snapshot of
the abstract? In the `real` paper, it makes sense, because the
reader will have access to the rest of the paper.
|
|
Boud's suggestions in the previous commit were great and really helped in
improving the tone of the abstract (and thus the whole paper shortly!),
better putting it in the big picture. I had forgot to give the exact word
limit (which was 250), so Boud had set it to a very conservative value of
190, I added around 22 words to better highlight the points we want to
make, while still being below the limit.
|
|
To make this a research article, we either have to present it as a
theoretical advance, or as an empirical advance.
An empirical research result would be something like doing a survey of
users and getting statistics of their success/failure in using the system,
and of whether their experience is consistent with the claimed properties
and principles of Maneage (e.g. success/failure in creating paper.pdf as
expected? was the user's system POSIX? did the user do the install with
non-root privileges? was this a with-network or without-network ./project
configure ?) This is doable, but would require a bit of extra work that we
are not necessarily motivated to do or have the time to do right now.
I think it's possible to present Maneage as a theoretical advance, but it
has to be worded properly. Maneage is a tool, but it's a tool that
satisfies what we can reasonably present as a unique theoretical proposal.
Here's my proposed rewrite. I've aimed at minimum word length. I've also
included (commented out) keywords for a structured research abstract -
these are just for us, as a guideline to improve the abstract.
I think "criteria" is safer than "standards". Whether a principle is good
or bad tends to lead to debate. Whether a criterion is satisfied or not is
a more objective question, independent of whether you agree with the
criterion or not.
In the rewrite below, we propose a theoretical standard and show that the
new standard can be satisfied. Maneage is *used as a tool* to prove that
the standard is not too difficult to achieve. Maneage is no longer the
subject of the paper. (That won't change the main body of the paper too
much, apart from compression, but the way it's presented will have to
change, under this proposal.)
The title would need to match this. E.g.
TITLE.1: Evidence that a higher standard of reproducibility criteria is
attainable
TITLE.2: Evidence that a rigorous standard of reproducibility criteria is
attainable
TITLE.3: Towards a more rigorous standard of reproducibility criteria
I would probably go for TITLE.3.
|
|
This abstract is a first step in order to put more focus on the research
aspects of Maneage.
|
|
Given the very strict limits of journals, we needed to remove these
sections and images. The removed images are: the
`figure-file-architecture', `figure-src-topmake' and
`figure-src-inputconf'. In total, with `wc' we now have 9019 words.
This will be futher reduced when we remove all the technical parts of the
Maneage section, in short, we will only describe the generalities, not any
specific details.
|
|
David suggested some interesting references in particular about the
problems with Juypyter notebooks that are now added to the long version of
the paper. We'll later decide if/how they can be used.
|
