Age | Commit message (Collapse) | Author | Files | Lines |
|
Split from mana/plus!79
|
|
Squashed with:
* More quoting, but these are likely unnecessary
[^"x]$[a-zA-Z]
[^"]x$[a-zA-Z]
****
mana/plus!98
|
|
Shaves some bytes from exe, too
****
mana/plus!102
|
|
TODO: are all these casts actually required? SMH
****
mana/plus!101
|
|
****
mana/plus!96
|
|
Squashed with:
* Appease linters + more reformat
* Further improve skill error handling and messages.
24751878 "Improve skill error messages." added trailing dots only on the
second branch. This adds missing dots on the first one as well as
compacts the logic.
****
mana/plus!93
|
|
& shaves ~5KB from debug binary.
previd=""; while read -r line; do if [[ $line =~ .*"msg.read"[^\"]*"\""([^\"]*).* ]]; then id="${BASH_REMATCH[1]}"; if [[ "$id" == "$previd" ]]; then printf "%s\n" "$line"; fi; previd="$id"; fi; done < <(grep -r 'msg.read.*' src/net/)
****
mana/plus!91
|
|
* Run gcc8 job again in this MR
* Squash the %c strftime gcc warning
****
mana/plus!100
|
|
Server does not communicate sent time.. or it did, until 2017 ??
See src/map/clif.c #if PACKETVER >= 20170419
The RODEX_EXPIRY constant seems not to be communicated either,
preventing us from deducing send time from expiry time.
****
mana/plus!90
|
|
if any, are worth it.
In fact, this might even save some time. Calculating distance may be
faster than iterating through memory twice.
****
mana/plus!95
|
|
for obsolete platforms. Move some others to only master...
Sunset/disable some CI jobs to reduce CI load. Move some others to only master branch as I deem them unlikely to fail if others work.
gcc7 is direct replaced by gcc8 (both really obsolete these days)
gcc7-default -> gcc10-default (same)
gcc10 game/dyecmd only ... no need to test if these two special build
options work on every ommit
Some special builders are made master-only pending further review.
Squashed with:
* Fix: only: has to be an array or ..
* Sunset/disable some CI jobs to reduce CI load. Move some others to only master branch as I deem them unlikely to fail if others work.
gcc7 is direct replaced by gcc8 (both really obsolete these days)
gcc7-default -> gcc10-default (same)
gcc10 game/dyecmd only ... no need to test if these two special build
options work on every ommit
Some special builders are made master-only pending further review.
****
mana/plus!99
|
|
****
mana/plus!94
|
|
Reported by Skyflare
****
mana/plus!97
|
|
How to reproduce: spam emote then quickly switch char
Thread 1 "manaplus" received signal SIGSEGV, Segmentation fault.
SkillDialog::getSkill (this=0x0, id=id@entry=1) at gui/windows/skilldialog.cpp:805
0 SkillDialog::getSkill (this=0x0, id=id@entry=1) at gui/windows/skilldialog.cpp:805
1 EAthena::SkillRecv::processSkillFailed (msg=...) at net/eathena/skillrecv.cpp:302
...
****
mana/plus!92
|
|
g
****
mana/plus!77
|
|
****
mana/plus!88
|
|
|
|
- [x] integrate docker-windows-builder with or instead of the existing disabled MXE jobs
- [x] fix rebranding breaking windows CI (already pushed to master)
Part of #71
Squashed with:
* $CI_COMMIT_BRANCH is unset in merge request pipelines
* Too much text
* Cleanup now that the builder branch is merged
* Revert "Disable unnecessary builders"
This reverts commit f4820e05439792a02601dfef0e2e2ad148615f1f.
* Try depend strategy + de-hardcode stuff
cleanup
* Use downstream branch created for this purpose (testring only, switch to
master before merging)
* Multi-project pipeline, attempt #1
missing stage definition
* Disable back, comment why it doesen't work
Forgot to disable the broken windows builder job
* Does existing MXE job even work? Kinda doubt it, but after this you
can't blame me for not trying
(What's the purpose of windows-docker-builder if this works?)
But lets try the *32 one as well
* Disable unnecessary builders
****
mana/plus!87
|
|
|
|
raise default (5->100).
|
|
|
|
Addendum to cbab9d6d "Fix for login button"
See also mana/plus!67 and mana/plus#64
****
mana/plus!86
|
|
every time'
No error/warning? odd.
|
|
This has the side effect that now #Debug history is also loaded.
****
mana/plus!84
|
|
Controller improvements
See merge request mana/plus!52
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
****
mana/plus!83
|
|
Saves 1min10s by not compiling and another 15s by moving it together with linters.
Squashed with:
* Make sure mplint internal error or failure to run is detected
* Make mplint breaking cause job to fail
* Move linters out of lightweight.. goes faster, 3m46s -> 1m45s
* Make script fail if any command returns nonzero or if any variable is
unset.
* Donwload mplint instead of building it
****
mana/plus!80
|
|
What happened:
1) herc evol sends 0x100 as packet ID offset
2) 0x100 is added to 0x0062 of SMSG_CHAR_PASSWORD_RESPONSE
3) SMSG_CHAR_PASSWORD_RESPONSE is then inserted into 0x162
slot of packet map, shadowing SMSG_GUILD_SKILL_INFO
4) on guild lvl-up, server sends this in SMSG_GUILD_SKILL_INFO
5) ManaVerse exits in confusion.
Note 1: due to popupmenu paste not working in change password dialog, tested password change only with invalid old password, which worked (opened error dialog instead of client exiting).
Note 2: Some packet handlers are bound using &Ea::LoginRecv prefix, this one had &LoginRecv and this was kept this way (uses vtable lookup?).
~~TODO: SMSG_CHAR_PASSWORD_RESPONSE seems to be something about password changes, but I could not find this packet either in TMWA or herc. Still, I need to test if these still work on all active servers~~ Tested.
Squashed with:
* Fix exit on password change, too
* Fix guild lvl-up packet parsing causing client exit in ML
****
mana/plus!81
|
|
Problem happens only on Windows that uses \ for path separators and only when primary download location does not contain requested file with requested checksum.
To repro, it should be enough to change one checksum in resources.xml
Report from Manatauro:
![pathjoin](/uploads/4b969168ba7babc21108c1db1a680f66/pathjoin.jpg)
****
mana/plus!82
|
|
|
|
Should this function gain emplace()-like semantics?
|
|
|
|
|
|
See #73 for discussion
Squashed with:
* Fix mana pearl showing in arrows slot in Classic
* Fix off-by-one. Was causing Old Towel to render in torso slot in ML.
****
mana/plus!78
|
|
See manaplus.log in https://git.themanaworld.org/specing/manaplus/-/jobs/168040#L3241
****
mana/plus!72
|
|
Screenshot code didn't take into account scaling setting which resulted in cropped screenshots.
I didn't touch SDL modes because they don't support scaling and I'm not sure how that would be implemented for software render.
****
mana/plus!70
|