fix(build): a bare "core" in .gitignore ate the core/ source tree - #42
Merged
Conversation
Root cause: .gitignore line 61 was a bare `core` rule - intended for core dumps,
but it matches the core/ source directory. Any new file under core/ was silently
dropped by `git add`. That is how core/common/include/eapps_core.h went missing:
it is included by 46 files across apps/ and exists in no commit in this
repository's history, so all 44 app targets failed at the first #include:
apps/ecal/ecal.h:5:10: fatal error: eapps_core.h: No such file or directory
The 42 files already under core/ are tracked (added before the rule bit), which
is why the breakage looked like a missing include path rather than a missing file.
Two changes:
1. `core` -> `core.[0-9]*` + `vgcore.*`, so core dumps are still ignored and the
source tree is not. Without this the header below cannot be committed at all.
2. core/common/include/eapps_core.h - an umbrella header. Nothing is invented;
every symbol the apps use was already declared elsewhere:
eapps_app_info_t, eapps_app_lifecycle_t <- eapps/types.h
eapps_registry_* <- eapps/registry.h
eapps_palette_t, eapps_theme_get_palette <- eapps/theme.h
eapps_scaffold_create, eapps_card_create,
eapps_list_create, eapps_list_item_create <- eapps/widgets.h
Verified with `ninja -k 0` on a clean configure: failing app targets went from
44/44 to 4/44. The remaining four are unrelated pre-existing defects, NOT fixed
here - each needs a decision about intended types:
ebot unknown type name 'ebot_state_t'
eremote 'MAX_DEVICES' undeclared; 'eremote_device_t' lacks 'volume'/'channel'
etrack 'LV_SYMBOL_SHOPPING_CART', 'lv_font_montserrat_12' undeclared
ewifi unknown type name 'ewifi_risk_level_t'; 'lv_font_montserrat_12' undeclared
etrack and ewifi are LVGL configuration (fonts/symbols not enabled in
extern/lv_conf.h), not application logic.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Verified: ninja exits 0, 0 FAILED, and all 44 add_subdirectory'd app targets
produce objects. Was 4 failing after the eapps_core.h fix (44 before it).
Two apps were TWO IMPLEMENTATIONS CONCATENATED, which produced the "duplicate
symbol" and "missing member" errors:
etrack.c A complete package-tracking app (lines 1-85: carriers, list view,
add view, full lifecycle) with an unrelated placeholder stub
appended after it - its own #include "etrack.h", its own
etrack_ctx_t, and duplicate etrack_init/deinit/info/lifecycle. The
stub shared no symbols with the real app. Removed the stub.
eremote.c Worse: the stub was pasted OVER THE TAIL of the real
eremote_init(), which lost its closing `return true; }` - the
function simply stopped after build_nav_bar(body). Restored the
ending, removed the duplicate init and ctx, kept the appended
lifecycle block because the real implementation defined none. Its
ebot_info metadata described a different app ("Remote desktop
client", EAPPS_CAT_NETWORK); corrected to the IR/WiFi remote this
actually is.
Two invented LVGL identifiers. LVGL defines 62 LV_SYMBOL_* and neither of these
is among them:
LV_SYMBOL_SHOPPING_CART apps/etrack -> LV_SYMBOL_LIST
LV_SYMBOL_DEGREE_SIGN apps/eremote -> "%d C" (the built-in montserrat
faces have no degree glyph, so a literal would have
rendered as a placeholder box)
Also enabled LV_FONT_MONTSERRAT_12 in extern/lv_conf.h: etrack and ewifi both
use it and it was not among the enabled sizes.
ewifi:
- `strlen(creds[i].security)` and `"%s"` on it - .security is an
ewifi_security_t ENUM, so this was pointer-from-integer. ewifi_security_str()
is the accessor and is already used correctly at :147.
- Its appended lifecycle block referenced an undefined `ctx`; the
ewifi_ctx_t/ctx definition had been lost. Restored from its uses.
FLAGGED IN A COMMENT, not fixed: that lifecycle is still a placeholder that
draws a title, status line and button and never calls the real UI in the same
file (build_scan_tab, build_channel_tab, build_security_tab,
build_passwords_tab, build_nav). Wiring it up is a product decision about tab
scaffolding, not a build fix.
eremote.h: added MAX_DEVICES (following the existing MAX_LEARNED/MAX_SCENES
pattern) and the device-state fields the real implementation sets - power_on,
volume, channel, temperature, fan_speed.
ebot.h: the header declared only the two eapps_* lifecycle symbols while ebot.c
used 17 undefined identifiers - the same failure mode as the missing
eapps_core.h. Recovered the type layer. Every struct field and width is derived
from ebot.c and the derivation is cited per-field in the header, e.g.
host[256] from `strncpy(state->host, ..., 255)`, models[].tier[16] from
`strncpy(..., 15); tier[15] = 0`. The five CAPACITIES (EBOT_MAX_MESSAGES,
_MSG_LEN, _MAX_MODELS, _MAX_TOOLS, _RESPONSE_BUF) are marked in the header as
choices, because nothing in ebot.c pins them; only EBOT_MAX_MESSAGES has
structural meaning, as the message ring shifts by one at :113-117.
Note on verification: an intermediate run of this work reported "0 failing"
because the CMake configure had failed on the LVGL FetchContent step and ninja
produced no output at all - not because everything built. The numbers above come
from a build tree with LVGL already populated, cross-checked by counting object
files per app target.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
This was referenced Aug 31, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
masterdoes not build. From a freshgit clone --depth 1:Every one of the 44 app targets includes
eapps_core.h, and it is not in therepository:
Cause
.gitignore:61A bare
corematches any file or directory namedcore, at any depth — soit matched
core/, the source tree. The rule was clearly meant for Unix coredumps; it silently excluded the shared headers instead.
Only the 42 files under
core/that predate the rule are tracked. Everythingadded since is invisible:
git statusshows nothing,git adddoes nothingwithout
-f, and the developer who wroteeapps_core.hhad a working buildlocally while
masterhad none.This is the second instance of this trap in the organisation. ebuild's
build/rule matched
ebuild/build/, its backend package, so a new module added therewas silently untracked — caught there before it cost anything, because nothing
had been added yet. Here it had already been paid.
The change
The rule now names what it was for:
plus
core/common/include/eapps_core.hitself, and repairs to the last four apptargets that had drifted while nobody could build them.
Verified
Both directions of the ignore rule, because loosening it must not stop it doing
its job:
Build from a clean tree:
Note for the open Dependabot PRs
#38, #39, #40 and #41 all show failing CI. That is this, not the action bumps —
their runs fail with the same
eapps_core.h: No such file or directory. All fourtarget versions do exist and are current (
actions/checkoutv7.0.1,actions/setup-javav6.0.0,pnpm/action-setupv6.0.10,mymindstorm/setup-emsdkv16), so once this lands their CI should be judgedagain on its own merits.