23

This one is only one of many examples were a user has spotted a bug or incompatibility in one of my LaTeX packages. A (more or less) dirty work-around or patch is provided as answer, and the user is fine with that.

Neither the user nor the person who provided the patch seem to think about reporting an issue to the package maintainer. Why?

Does no one care because it's a win-win-situation for both parties, the user has a solution and the one who gave the patch earns reputation?

If I take a look at the caption package bug tracker I mostly see bugs I have found on SX and have put it into the bug tracker myself. I must confess that I feel not motivated at all to fix them since no-one seem to care. And since I don't visit SX regularly these bugs are only the tip of the iceberg, I guess there are many more hidden in SX questions & answers I'm not aware of.

  • 8
    Very good question! I was asking me the same several times! This probably won't help you for the caption or subcaption tags, but for tags that are clearly associated with only a single package, I hope that including a link to the bug tracker in the tag info will help, e.g. tex.stackexchange.com/tags/tikzducks/info – user36296 Mar 15 at 20:37
  • Another idea: if there are any users who are particular active in tags related to the caption page make sure that they know that you prefer bug reports. They probably can judge reasonable good if something might be a bug and either make a bug report or remind others to do so. – user36296 Mar 15 at 21:20
  • 9
    Sadly, some developers fail to respond to even rather urgent bug reports for extended periods. (This is not directed at you; I am thinking of other offenders.) Unfortunately, that can have a discouraging effect on a user's willingness to go to the trouble to prepare a cogent bug report and submit it. I think your request is well presented, and I hope it bears fruit. – barbara beeton Mar 15 at 23:55
  • @barbarabeeton I'm definitely no exemplar of a "good maintainer", my contribution is fluctuating, the caption documentation is outdated etc. But I have a family and a full-time job, and have moved house six times since publishing the caption package. I see that circumstances like this may have a discouraging effect on the user, but how to solve this issue? In 2019 there are so many tools available for contributing to open-source stuff, bug trackers, wikis, pull requests etc., but this does not help if these tools are not used. – Axel Sommerfeldt Mar 16 at 11:35
  • 3
    IMHO the current situation is: SX is some kind of bug tracker for LaTeX packages, but the maintainers do not get notified about the problems. And since the users get work-arounds here, the amount of "real" bug reports have dropped significantly since tex.sx exists. So in fact SX is some kind of huge data base for bugs & work-arounds, but does not help to improve the quality of the packages. – Axel Sommerfeldt Mar 16 at 11:36
  • 2
    ...Which is a pity since SX saves time by doing package support. (I guess Ulrike has done more support of my packages than myself, and I'm really grateful for that.) And SX helps analyzing the problem, doing MWEs etc. SX ist really a great first-level-support, but the communication way to the maintainer is missing so unfortunately the product don't get fixed. – Axel Sommerfeldt Mar 16 at 11:43
  • 2
    @samcarter -- Very good suggestion to include a bug report destination in the tag text. I'm making a list ... – barbara beeton Mar 16 at 21:49
  • I don't think this is a bug of newfloat. It's a bug of mwrep. – egreg Mar 24 at 10:25
22

Let me give an answer from the point of view of a LaTeX user who is sufficiently expert to do most of the things they need but not enough expert to understand most of the underlying details.

If I look at the question and at Enrico's answer, I wouldn't be able to understand that that's a bug or that the incompatibility would be solvable by a bug report. So, I wouldn't even think to file a bug report.

And, starting from LaTeX 2.09, I've been using LaTeX for almost 25 years now. How can a common user really understand what's a bug and what's not unless explicitly told? The association from it doesn't work this way to it's a bug is not automatic.

Note also that many LaTeX users are not familiar at all with software development (something that many developers give for granted) and they are not aware of the possibility of filing bug reports at all.

To put it as one of my colleagues once told me after I solved one of their LaTeX problems: What could I have done without you? (which, at least, gives sense to my life :-p)

  • 2
    What about tex.stackexchange.com/questions/477824/… ? It even starts with "I found a issue". I agree that it's not easy for the user to decide if a bug report should be issued, but the provider of the patch can clearly do so. How to fill a bug report? I would be happy to get the link to the SX question via E-Mail (or via GitLab), this would be sufficient. – Axel Sommerfeldt Mar 15 at 19:42
  • 5
    @AxelSommerfeldt I agree with Massimo that it is hard to tell whether something is a bug. Plus some package author may feel criticized unfairly if something is referred to a bug which may also be seen as user error. I really like the fact that you really want to improve your packages and ask for feedback, but one should understand that there is a huge threshold for everyone who does not understand every detail of the package. – user121799 Mar 16 at 4:10
  • 2
    While this answer and comments give good inside for me what the problem is, I'm still looking for solutions. How to lower the threshold? – Axel Sommerfeldt Mar 16 at 11:24
  • 4
    @AxelSommerfeldt Questions I'd ask myself if I were the author of a package and wanted bug reports: (1) Is it easier to type up something and click on "Ask a question" on TeX.SE (or post on some LaTeX forum, etc), or is it easier to report a bug? (Easier to report bugs => Lower threshold) (2) If a user uses the package, how likely are they to know how/where/when to report bugs? (E.g. is it mentioned on the first page of the documentation?) (3) What is the likely outcome of reporting bugs, i.e. is a user more likely to get an answer (and more quickly) by posting on TeX.SE or on the bug tracker? – ShreevatsaR Mar 17 at 19:35
  • 4
    @AxelSommerfeldt Ultimately at some point your question becomes “how can I get users to do something that is harder and less likely to be useful to them, instead of something that is easier and more likely to solve their problem”, and there's really no answer to that. I think your current approach, of looking at various places on the internet where users might have discovered bugs, is good. (Maybe you can nudge question askers or answerers to post a link to the TeX.SE (or whatever) question; that's about the best achievable outcome but even that's extra effort, e.g. creating a Gitlab account.) – ShreevatsaR Mar 17 at 19:38
  • 2
    I think this answer and @ShreevatsaR’s comments really nail the point here. Other answers suggest lots of good things which may nudge users to submit more bug reports, and that’s great, but fundamentally we also need to accept that for very natural reasons, there will still be lots of people who ask here instead. – Peter LeFanu Lumsdaine Mar 19 at 10:39
  • 2
    While I really appreciate all the answers to my question (Thanks a lot, folks!), this one has clearly shown to me that there is no proper answer since encouraging the user to fill a bug report is simply the wrong approach. And since there is no CTAN bugtracker, TeX.SX is filling that gap. Unfortunately I cannot use TeX.SX as bugtracker since looking at all questions tagged "caption" is not an option, there are simply to much questions tagged with it and only a fraction is dealing with my package. – Axel Sommerfeldt Apr 13 at 19:14
17

If I see a question that indicates a bug, I normally write in my answer or in a comment "this is a bug, notify the maintainer".

I normally don't make the bug report myself as I think it is important that everyone learns to make bug reports. "You discovered the bug, you should also report it".

But if the bug affects an important current package (and I would count caption as such a package) I try to keep track if a bug report is made.

  • 4
    I take the opposite approach, generally. If I solve a problem that turns out to be a bug, I usually file a bug report myself. While I agree that people should learn how to make bug reports, I think those of us who figure them out are likely to make much more useful bug reports too. You, for example, are a master at constructing truly minimal examples, whereas someone who uncovers a bug may have asked a question with a much more elaborate example. – Alan Munn Mar 16 at 3:04
  • 3
    @AlanMunn how should they learn how to make bug reports if you tell them "let me do it I'm better at it"? Or if you show them that they don't need to care as some expert will do the work for them? – Ulrike Fischer Mar 16 at 11:36
  • 1
    Telling people to make a bug bug report isn't really teaching them how to do it. Furthermore, most people on the site just want to solve their problem, they're not invested (or they don't see themselves as invested) in the larger enterprise of community written software, independent of whether they feel competent to report the bug themselves. So I think it's helpful if those of us who are so invested report bugs we fix. – Alan Munn Mar 16 at 12:42
  • 2
    @AlanMunn the content of the bug report is not really the problem. After all they were able to describe the problem on tex.sx so that I could see that there is a bug and a link to tex.sx is often quite enough for the maintainer. The main hurdle is that they often don't dare and care to contact the author. – Ulrike Fischer Mar 16 at 12:57
  • 1
    Yes, I agree with that. So those of us who do dare and care can help out by doing the reporting as well. If people don't dare or care to report bugs, telling them they should report a bug not likely to have much effect. – Alan Munn Mar 16 at 13:00
  • 1
    @AlanMunn it works quite ok for me. Not everyone does, but quite a lot sent bug reports if I tell them to do it - sometimes starting with something like "Ulrike said it is a bug ..." to put the blame on me for the case that the report is nonsense ;-). – Ulrike Fischer Mar 16 at 13:08
  • @UlrikeFischer Maybe we can add to the comment a link to this question: tex.meta.stackexchange.com/questions/8170/…. – K-HB Mar 17 at 13:16
11

This answer summarizes my comments to Ulrike's answer.

When I solve a bug on the site, I usually file a bug report myself. While I agree that it is good to encourage people to file bug reports themselves, I think there various reasons why the average user on the site is not so likely to do it, from feeling incompetent to do so to not really caring much, since their immediate problem has been solved. For those of us who already use GitHub or the like, filing a bug report seems simple, but if you're not a computer person it's not so trivial. And emailing the author also requires some digging sometimes to get an email address. Filing a bug report has no direct benefit to the filer; the benefit is more abstract (it helps others) and long term, and these reasons alone probably account for why people don't file them.

Since the site has become a de facto bug reporting system, I think those of us who do solve bugs should help out by additionally filing a bug report. Of course if the questioner is independently an active answerer on the site who is likely to file the report themselves, it probably suffices to just suggest they do so, but in other cases, we can be most helpful by taking on the task ourselves.

  • 1
    +1 Sometimes the answer provided SX includes a way to fix the bug (rather than just bypassing it). In this case, it is better for the author of the answer to report the bug and the fix himself. – Paul Gaborit Mar 19 at 15:40
8

I don't think the problem is really solvable and has always been this way, before tex.sx there were far more bugs "reported" on comp.text.tex newsgroup than via the latex gnats bug tracker.

To take the example you give. You can't necessarily expect the original user to know where the bug is, or even if it is a bug not user-error. So asking on tex.sx is a reasonable course of action. Then given egreg's answer it still isn't clear that the bug should be reported on the newfloat bug tracker. It could be read as saying that the chapter test in newfloat is not generic enough and so an enhancement request to newfloat should be made, or it could be read as saying the class is violating some (poorly documented) convention that classes defining chapter level constructs should follow, and so a bug report should be made to the class.

The actual workaround posted would if anything favour the second reading, as it essentially patches the class to define \@chapter rather than patching newfloat, but that's just for ease of coding a workaround not necessarily suggesting a fix should be in either place.

It often works best if general user questions on forums such as stackexchange get converted to more focussed bug reports by the developer themselves, for longtable for example I try to track reports from various places (with mixed success)

If you don't follow the site regularly you could still get it to alert you with any questions tagged newfloat

https://stackexchange.com/filters/364213/my-filter

or perhaps even just tagged newfloat and bug, then people would just need to use those tags to alert you, rather than needing a github account.

  • That's a good idea, but unfortunately the newfloat label will also be used for problems with \newfloat offered by the float package. When looking at caption is even gets worse. So what is missing are package-related labels I can subscribe to, for example package-caption or package-newfloat. – Axel Sommerfeldt Mar 17 at 14:42
  • 1
    @AxelSommerfeldt also probably more than half of questions are tagged incorrectly originally, it's not ideal but it is what it is... – David Carlisle Mar 17 at 14:47
  • Why "it is what it is..."? This could be changed by introducing new package related labels, couldn't it? – Axel Sommerfeldt Mar 17 at 20:16
  • @AxelSommerfeldt well I was mostly thinking of the fact that people will always mis-tag. the [newfloat] tag could be specified to be just for that package, currently it has no definition at all, see tex.stackexchange.com/tags/newfloat/info but even if the tag wiki was filled in, tagging is very hit and miss in terms of accuracy to be honest – David Carlisle Mar 17 at 20:56
  • @AxelSommerfeldt I think you have a good point that package specific tags that are clearly marked like caption-pkg or package-caption as you suggest might be helpful, especially for those packages like yours where the name overlaps substantially with the topic. They might even reduce somewhat the bad tagging, although David is correct in saying that tagging (especially initial tagging) is really hit and miss. – Alan Munn Mar 17 at 23:20
  • 1
    @AxelSommerfeldt There's also precedent for this method: we have both a cite-package and a proof-package tag, so in principle we could create more, although for things like caption the retagging effort would be enormous and potentially disruptive, since retagging an old question puts it on the front page. This can be mitigated somewhat by retagging in small batches over a period of time. – Alan Munn Mar 17 at 23:32
4
  • First the user has to find out, if it is only the own system, or a package bug. It is not trivial to test TeX packages: Test routines are often missing.

  • Most TeX users have no clue about packages and TeX internals. This is different for other areas. Someone who uses a python module can usually read its source well and write a bug report easily.

  • It should be easier for a user to file a bug in TeX packages. For the average "part time TeX user" it can take hours to find the current website, repository and mail contact of the broken package. CTAN could force stricter package formats including this information in a human and computer readable format.

  • Many packages are badly maintained or not maintained at all. I think the motivation to invest hours in a proper bug report, for a package without any updates in the last years is quite low for most users.

1

Maybe one could build a system where all questions tagged are fed into CTAN's bug tracking system.

  • 8
    AFAIK there is no CTAN bug tracking system. Each package maintainer is responsible for tracking their own bugs. Some use 'modern' methods like GitLab or GitHub, many prefer simple emails. A not insignificant number of packages is factually unmaintained. So the first step here would be to turn CTAN from a simple dissemination platform into a development platform. I'm not sure if that could be pulled off easily with the current resources (personal as well as technical). – moewe Mar 26 at 7:32

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .