ID |
Date |
Icon |
Author |
Author Email |
Category |
OS |
ELOG Version |
Subject |
69738
|
Thu Feb 22 08:12:28 2024 |
| Stefan Ritt | stefan.ritt@psi.ch | Question | Linux | Other | Latest version | Re: no availability of el8 and el9 rpm |
> I spent one hour yesterday to install almalinux in VM on my laptop, clone repo, rebuild ELOG RPM, install the new rpm and start elogd.
> You should have several VM on a shield for that :-P
Right. One hour to install linux X, another hour for linux Y. Building RPMs is a constant challenge taking lots of time. If we
find someone who volunteers to build the RPMs and EXEs from the sources and takes over the responsibility for the next years,
I would be more than happy to upload the resulting files. I simply don't have time for that given all my other responsibilities.
Best,
Stefan |
69740
|
Fri Feb 23 08:40:26 2024 |
| Stefan Ritt | stefan.ritt@psi.ch | Question | Linux | Other | Latest version | Re: no availability of el8 and el9 rpm |
> For my curiosity, I've just clone ELOG repo on Bitbucket to try the process. That's pity : pipelines are currently broken (incident traced in status page https://status.atlassian.com/).
> The goal is to build elog and pkg file for Ubuntu (docker images list : https://status.atlassian.com/). Next steps should to create a AlmaLinux VM to try RPM builds.
I fixed the bitbucket pipeline. I compiled it for an (old) GCC 6.3, and even if I installed cmake there, it was not running (maybe I needed cmake3?). Anyhow, I switched GCC 10.2 and now it's fine.
> With BitBucket, it's possible to configure personal runners for CI/CD (systems with pipeline client installed, as RHEL, Fedora, Windows, etc...). The pipeline scripts can build, test and generate the packages automatically, depend on system runners.
> As CI/CD, the process is automatically done after commits in git repo... A lot of work to make it in place, but when it's running, the delivery process is done
> This is also possible with GitLab and other big platorms.
As you saw I use the bitbucket pipeline for CI/CD. The current bitbucket-pipelines.yml is pretty simple (see attachment). If you make me one which does the rpmbuild automatically, I'm more than happy to upload and use it. The downside there is that it only works for so long. As you saw at the
error above, the pipeline worked a few years ago when I installed it. But in meantime things changed apparently and need to be fixed. ELOG is around since last century (literally!), and I'm kind of tired to fix things every once in a while. If somebody else could take over the responsibility to
deliver the RPMs I would be more than delighted.
Stefan |
Attachment 1: bitbucket-pipelines.yml
|
# Build ELOG with GCC 10.2 and latest
image: gcc:10.2
pipelines:
default:
- step:
name: Build on GCC 10.2
image: gcc:10.2
script:
- apt-get update && apt-get install -y imagemagick cmake
- git submodule update --init
- mkdir build; cd build
- cmake ..
- make
- step:
name: Build on GCC latest
image: gcc:latest
script:
- gcc --version
- apt-get -qq update && apt-get -y --force-yes -qq install cmake
- git submodule update --init
- mkdir build; cd build
- cmake ..
- make
|
69741
|
Fri Feb 23 14:59:29 2024 |
| Stefan Ritt | stefan.ritt@psi.ch | Bug report | Linux | ELOG V3.1.5 | Re: http status 200 returned for "file not found" |
> > "file not found" should return http code 404. elogd returns code 200 together
> > with a page containing text "404 not found". This pollutes the browser cache
> > with wrong content (in this case, we are trying to load a css file, and the browser
> > is trying to use text "404 not found" as if it were a css. bad. file not found
> > should return http code 404. K.O.
>
> Yes. That's quite a problem when interacting with ELOG programmatically. Only way to
> find whether response succeeded or failed with 404 is to parse response body
>
> When file is not found send_file_direct calls show_html_header which in turn calls
> show_http_header which sets HTTP code 200 unconditionally. It's reasonably easy to
> patch around.
I fixed the code to properly return "404 Not Found" in case a non-existing file is requested.
Stefan |
69742
|
Fri Feb 23 15:13:46 2024 |
| Stefan Ritt | Thstefan.ritt@psi.ch | Bug report | Linux | 3.1.5-1 | Re: user change under webserver authentication not recognized |
Thanks for the fix, I committed it. Please give it a quicky try sinc I cannot test it here (don't use webserver authentication...)
Stefan
Frank Heyroth wrote: |
I found the reason of the bug:
In line 27441 of elogd.cxx the http_user is overwritten by the user saved in the sid_ array as a sideeffect of the sid_check function:
sid_check(getparam("sid"), http_user)
It can solved by changing elogd.cxx @ line 27441
27441c27441,27446
< if (!sid_check(getparam("sid"), http_user)) { /* if we don't have a sid yet, set it */
---
> i=sid_check(getparam("sid"), thumb_name);
> if (i && strcmp(http_user,thumb_name)!=0) { /* user changed */
> sid_remove(getparam("sid"));
> i=FALSE;
> }
> if (!i) { /* if we don't have a sid yet, set it */
Remark: I have used the variables i & thumb_name of the function in a local context.
|
|
69744
|
Mon Feb 26 17:39:56 2024 |
| Stefan Ritt | Mstefan.ritt@psi.ch | Question | Linux | Other | Latest version | Re: no availability of el8 and el9 rpm |
Many thanks for your files and instructions. I integrated this into the official elog pipeline, so now the RPMs can be downloaded from https://elog.psi.ch/elog/download.html and they will be updated automatically after each commit. If you succeed with any Debian/Ubunto I'm happy to add that. What's missing now is an easy way to compile for Windows (which I don't have anymore).
Stefan |
69748
|
Tue Feb 27 07:35:15 2024 |
| Stefan Ritt | stefan.ritt@psi.ch | Question | Linux | Other | Latest version | Re: no availability of el8 and el9 rpm |
Laurent Jean-Rigaud wrote: |
For Windows target, It's supported in Bitbucket but only with private agent :-(
So I tried to build using docker hub MSVC image (abrarov/msvc-2019) but pipeline fails to retrieve the image. Maybe I need to swipe on alternative images, or these images with MS software are unaccepted from Atlassian...
Also, I tried to crosscompiling ELOG Win32 from Linux with g++-mingw-w64-x86-64, but Makefile/sources need to be patched to manage this 'platform'... For the moment, it fails. :-(
Next solution : use Wine in Debian image to build Elog with mingwin as I know that it works (under Windows :-)). Will be tested after all...
|
For Windows it's more than just getting the .exe. Users want an installer. I used the Nullsoft Scriptable Installer (NSI), but I have no clue if this still works today. But there is still the elog.nsi script in the repository. Next, users want elogd as a service running in the background as a "service". I'm pretty sure the old way of doing that has changed and one would at least have to test this. Unfortunatley I switched to MacOSX about ten years ago (actually most of ELOG has been developed under Windows, but these days are gone), so I'm kind of illiterate in the Windows world now and don't have any time to change that.
Stefan
|
69756
|
Sat Mar 9 17:14:09 2024 |
| Stefan Ritt | stefan.ritt@psi.ch | Bug report | Mac OSX | 3.1.5-23df00d9 | Re: Runaway bogus attachment counts in Summary view attachment column |
I'm traveling right now and will only next week be able to look into that, so please be patient for a few more days.
Stefan |
69761
|
Thu Mar 14 21:21:33 2024 |
| Stefan Ritt | stefan.ritt@psi.ch | Bug report | Mac OSX | 3.1.5-23df00d9 | Re: Runaway bogus attachment counts in Summary view attachment column |
The problem with the attachment paperclips has been fixed in the current version.
Stefan
Andreas Warburton wrote: |
On a new MacBook Pro (Silicon M3), I installed version 3.1.5 build 23df00d9. The application appears to work normally, except that, after a short while, the indicated attachment count (paperclips in the attachment column of the Summary view) starts to increment fairly rapidly with each time that I visit the page. Attachment counts appear even for records that don't have any attachments. When I access records individually, either those with or without real attachments, everything looks OK. Any insights as to what might be causing this, and how to correct?
|
|