When shell_exec() is disabled on the server (common in production), the system
operations panel now gracefully falls back to PHP built-in functions:
- gethostname() for hostname instead of shell 'hostname' command
- php_uname() for OS info instead of shell 'uname -a' command
- /proc/uptime parsing or unavailable status for uptime
- disk_free_space() and disk_total_space() for disk usage
- memory_get_usage() and PHP memory limit for memory usage
- Improved parsePhpMemoryLimit() to handle memory limit parsing
This ensures the admin panel shows meaningful information even in restricted
production environments where shell_exec is disabled for security reasons.
- Added System Operations link to Admin dashboard (app/Views/Setting/Admin/index.php)
- Uses load_form_div to load the system operations page inline
- Added SystemOperations library for server status collection and actions
- Tile uses server icon (bi-server) and appears in the admin panel grid
- Link: javascript:load_form_div(base_url('setting/admin/system-ops'), 'maindiv', 'System Operations')
- Added SystemOps controller with index and panel endpoints for dynamic content refresh
- Created system_ops_panel.php fragment for AJAX-based UI updates (no full page reload)
- Implemented 30-second auto-refresh using AJAX instead of location.reload()
- Fixed health badge contrast by adding explicit dark text color (#1f2937)
- Added jQuery availability checks to prevent script errors on page load
- Included server overview, resource usage, services, events, and update history
- Integrated delegated event handlers for action buttons (Update, Restart, Reboot, Shutdown)
- All buttons are protected with confirmation dialogs
- Layout maintains NiceAdmin card styling and responsive design
Routes added:
- GET /admin/system-ops/panel - returns panel fragment for refresh
- Diagnosis.php: resolve doctor mapping via ABDM Report Doctor Mapping,
compile pathology/diagnosis reports into separate Letter Head and
Plain Paper PDFs, add compiled-pdf-links endpoint
- Routes.php: add diagnosis/compiled-pdf-links route
- pathology_detail.php: render/persist dual compiled PDF buttons
- Setting/Template.php: per-modality Letter Head / Plain Paper template
mapping via hospital_setting keys
- Setting/Template/index.php: fix tile links using load_form (which only
replaces #main) instead of load_form_div, which broke every internal
AJAX reload (template type switch, modality switch, reset) on the
diagnosis/discharge/document print template settings pages
- diagnosis_print_settings.php: fix corrupted watermark_alpha number
input (missing step attribute) that silently blocked form submission
for templates with non-integer-step legacy values
- Patient.php: broaden non-fatal-pending detection in abdm_content_auto_flow() to recognize more bridge error wordings (No HIU sessions found / matching the criteria) and HTTP 400, stopping spurious 'Auto flow failed' toasts.
- Opd_prescription.php: remove dead opd.fhir.generated bridge_sync_queue enqueue in storePrescriptionFhirBundle(). This unprefixed event type routed to the CSNOtk terminology bridge which has no consumer for it, always 404ing. Real ABDM sharing already works via DreamsoftConnector::sharePrescriptionBundle() (abdm.fhir.share.requested).
- Fix groupWorkflowRowsIntoSessions() to correlate workflow rows by stable
request_id/abdm_consent_request_id instead of chronological proximity to
the newest CONSENT_REQUEST row. Background polling reconciles all
historical sessions in one batch, so older sessions' rows previously got
silently merged into the newest session, causing 'Show Data' to display
an unrelated older session's single facility instead of all consented
facilities.
- Skip synthetic orphan sessions (stray rows with no owning CONSENT_REQUEST)
in getAbdmConsentRequestsList()/getAbdmConsentArtifactDetail() so they
don't show as phantom history rows or get picked as 'most recent'.
- Fix abdm_documents() consent_request_id filter to match all known sibling
artifact IDs for a session (via abdm_hiu_consent_artifacts), not just an
exact single artifact id, so all facilities under one consent request
appear together.
- Fix CreateAbdmHiuConsentArtifacts migration: reduce VARCHAR(180) to
VARCHAR(100) on consent_request_id/artifact_id to satisfy MySQL's
1000-byte composite key length limit (was silently failing to migrate).
- Add fetchAllArtifactsAfterGrant() cascade in M3HiuWorkflowService to fetch
every known artifact under a granted consent, used by both the cron poll
and a new abdm_check_live_status() endpoint/button.
- getLatestAbdmSyncSnapshot() session-scoping fix to avoid an older granted
session masking a newer in-progress one.
- Relabel Dr. Joshi Clinic documents (previously mislabeled 'Test Hospital'
from the HIP's own FHIR bundle Organization name).
Abha::verifyOtp() and Abha::verifyCommOtp() were overwriting the
user-typed, OTP-verified mobile number with whatever the ABDM bridge's
verify-otp response echoed back for 'mobile'. Many bridge responses
(especially plain mobile-OTP verification, as opposed to Aadhaar eKYC)
do not echo the mobile at all, so it silently became empty and was
never saved to patient_master.mphone1 during ABHA Register (By ABHA
Number / QR / Facility QR / Mobile OTP) or ABHA Create/Verify.
Fix: fall back to the originally submitted/verified mobile when the
gateway response omits it. Also updated Patient_V.php JS (Verify ABHA
tab, ABHA Register Method 1 and Method 4) to resend the mobile number
along with the verify_comm_otp call so the server has it available.
Address/email are unaffected by this bug: those fields are already
mapped via applyAbhaMetaColumns() when present, but the ABDM mobile-OTP
verification API generally doesn't return KYC demographic data (that
only comes from the Aadhaar-eKYC flow) - so they may legitimately
stay blank for phone-based registration methods.
HMS instances behind NAT with no live public IP can hit an expired/unreachable
TLS cert on their own base_url() when doing the internal handoff self-call to
AbdmGateway::share_prescription_bundle, causing 'Gateway handoff failed:
curl error 60: Peer's Certificate has expired'. This call never leaves the
local server, so skip peer verification by default via new
fhirGatewayHandoffSslVerify() helper. Set FHIR_HANDOFF_SSL_VERIFY=true in .env
to re-enable strict verification on deployments with a valid public cert.
Does not affect the actual outbound ABDM bridge submission (EAtriaBridgeConnector),
which keeps its own independent SSL verification/toggle.
Those operations can store the entire decrypted FHIR bundle (incl. base64 scanned-document attachments, sometimes MB per row) in response_json. fetchAbdmWorkflowRows()/getLatestAbdmSyncSnapshot() only ever read small consent-status/hi_types fields out of response_json, which never appear in bulk data_fetch payloads anyway, so cap it at the SQL level instead of buffering it into PHP memory just to build a list view.
Confirmed empirically: the SAME clinical document (identical FHIR Bundle.id) gets re-delivered by the bridge under a DIFFERENT transaction_id/consent_ref every time it is re-fetched (e.g. via the by-ABHA-address historical fetch, which returns every past consent session for the patient). The old dedup hash included transaction_id+consent_ref, so every re-fetch created a brand-new duplicate row instead of updating the existing one -- observed as literal duplicate rows in the OPD ABDM Fetched Data list (same date/title appearing 2-3x). bundle_id is set once by the HIP and is stable across re-fetches/consents, so pair it with care_context_reference to uniquely identify a document instead.
- M3HiuWorkflowService: data_fetch now supports fetch_by_request_id=1, sending request_id (not consent_id) per bridge Method 1 -- targets the current/self consent request's own session, distinct from Method 2 (consent_id, which can resolve to a stale session sharing the same artifact).
- Patient::abdm_content_fetch_only(): tries Method 2 (consent_id) -> Method 1 (request_id, self data) -> Method 3 (abha_address, historical fallback) in order, stopping as soon as real documents are found.
- Fixed a real pre-existing bug this surfaced: getLatestAbdmSyncSnapshot() used the same flawed 'status=SUCCESS implies COMPLETED' check (without requiring workflow_state=DATA_RECEIVED) already fixed elsewhere in computeConsentSessionDetail(). This let broad/historical data_fetch rows (status=success, but decrypted_data=[]) hijack priority-500 'best' selection and silently drift the resolved 'latest granted consent' to an unrelated historical consent_id after Method 3 fallback ran. Now requires workflow_state=DATA_RECEIVED, matching the earlier fix.
Per bridge gateway team guidance, GET /v1/hiu/data/fetch?abha_address=... returns ALL historical decrypted records across every past consent for a patient, unlike the consent/request-scoped Methods 1/2 previously used exclusively. Added M3HiuWorkflowService support for a fetch_by_abha_address flag on the data_fetch operation, and wired abdm_content_fetch_only() to automatically fall back to this broader lookup when the consent-scoped fetch finds nothing new -- this surfaced 14 previously-invisible historical documents (consultation reports, pathology/cytology) for the test patient that were never being fetched before.
abdm_content_fetch_only() always returned phase=COMPLETED and 'Records fetched successfully' whenever the data_fetch API call itself succeeded (fetchResult.ok===1), even when documents_persisted+documents_updated was 0 (bridge/HIP hasn't actually delivered any records yet -- decrypted_data still empty, session status 'requested'). Now check documents_persisted+documents_updated: if 0, report phase=GRANTED with a clear 'not sent any records yet' message instead of falsely claiming success.
computeConsentSessionDetail() treated any data_fetch/hi_data_push_callback row with status=success as proof the session reached COMPLETED. But the DB 'status' column only reflects whether the underlying API call/poll succeeded (no HTTP/transport error) -- NOT whether data was actually received; that's tracked separately via workflow_state (DATA_PENDING vs DATA_RECEIVED). A data_fetch poll that succeeds at the HTTP level but returns decrypted_data=[] (bridge still processing) was incorrectly marked COMPLETED instead of staying at GRANTED. Now also require workflow_state === DATA_RECEIVED for the COMPLETED phase.
When a consent session's overall phase resolves to GRANTED/COMPLETED via a data_fetch or hi_data_push_callback row (proof data was actually fetched), but no consent_reconcile/consent_status row in that session ever recorded which hi_types were granted (e.g. data arrives via async push before any reconcile poll runs), grantedHiTypes stayed empty and every requested HI type was incorrectly marked DENIED. Fall back to treating the full requested set as granted when we have positive proof of GRANTED/COMPLETED but no negative evidence of per-type denial.
